refactor(workflows): merge skills into the unified workflows primitive

[sanil-23]

Summary

• Unifies the legacy skills domain into a single workflows primitive: src/openhuman/skills/ → src/openhuman/workflows/, with the RPC surface, agent tools, type idents and FE renamed in lockstep (skills_* → workflows_*, run_skill → run_workflow, Skill* → Workflow*).
• Dissolves the dead agent_workflows domain + phase machinery (workflow_load / workflow_phase tools removed); the unified create form combines skill + workflow (adds when_to_use).
• Removes the bundled default skills and the retired Runners/scheduled-jobs FE; workflows now live on the Intelligence surface only.
• Adds the run_workflow execution + discovery tool family (list_workflows, describe_workflow, list_workflow_runs, read_workflow_run_log, run_workflow, await_workflow) wired into the orchestrator.
• e2e coverage for the workflow create→run path, the task→workflow spine, and run-spawn guards.

Demo

🎥 Loom walkthrough

Problem

The product carried two overlapping primitives — "skills" and "agent workflows" — with parallel RPC namespaces, FE surfaces and agent tools. This split confused the create/run UX and left dead machinery (the QuickJS-era runner, phase tools, scheduled-jobs overview) on the tree.

Solution

Collapse everything onto one workflows domain. The rename is mechanical-but-wide (Rust core + RPC wire keys + FE callers + i18n), staged as discrete commits (domain dir → RPC methods → type idents → execution path → FE → cleanup) so each step is reviewable. The run-path loader is unified with discovery so created workflows are immediately runnable.

Submission Checklist

• Tests added or updated — workflow create→run e2e, task-spine e2e, run_workflow spawn-guard tests
• N/A: Diff coverage ≥ 80% — the diff-cover Coverage Gate is skipping on this PR; consolidation/rename of existing covered code.
• N/A: Coverage matrix updated — no new feature rows (rename/consolidation of existing skills features).
• All affected feature IDs listed under ## Related — N/A (no new matrix IDs)
• No new external network dependencies introduced
• N/A: Manual smoke checklist — no new release-cut surface (existing Intelligence/workflows surface).
• N/A: Linked issue closed via Closes #NNN — no tracking issue for this refactor.

Impact

• Desktop (core + Tauri + React). No mobile/CLI-specific behavior change beyond the shared RPC rename.
• Migration/compat: JSON-RPC methods and wire keys rename skills_*→workflows_* and skill_id→workflow_id; FE updated in lockstep. External callers of the old skills_* RPC names would need updating.
• Removes dead code paths (agent_workflows, phase tools, bundled default skills, Runners tab).

Related

• Closes:
• Follow-up PR(s)/TODOs: task-sources board work is intentionally not in this PR (kept on a separate branch); it stacks on top of this migration.

---

AI Authored PR Metadata

Linear Issue

• Key: N/A
• URL: N/A

Commit & Branch

• Branch: refactor/skills-workflows-merge
• Commit SHA: d000b5cd

Validation Run

• pnpm --filter openhuman-app format:check — prettier --check clean
• pnpm typecheck — tsc --noEmit clean
• Focused tests: cargo check --tests + cargo check --bin openhuman-core green against current upstream/main; i18n parity clean
• Rust fmt/check (if changed): cargo fmt --check clean; cargo check green
• N/A: Tauri fmt/check — no app/src-tauri changes in this PR.

Validation Blocked

• command: N/A
• error: N/A
• impact: N/A

Behavior Changes

• Intended behavior change: one unified "workflows" primitive replacing skills + agent_workflows.
• User-visible effect: skills surface/tabs consolidated onto Intelligence → Workflows; create form unified.

Parity Contract

• Legacy behavior preserved: workflow create/run/discover parity with prior skill run path; run-path loader unified with discovery.
• Guard/fallback/dispatch parity checks: run_workflow spawn guards (re-entrancy, await cap, backstop) covered by tests.

Duplicate / Superseded PR Handling

• Duplicate PR(s): N/A
• Canonical PR: this
• Resolution: N/A

Summary by CodeRabbit

• New Features

  • Intelligence gains a Workflows tab with deep-linking and new Create (/workflows/new) and Run (/workflows/run) routes.
  • Create workflow supports "When to use" and edit mode; installed workflows list offers Run and Uninstall actions.

• Improvements

  • Runner: locked-run mode, stop in-flight runs, auto-refreshes recent runs, and shows scheduled-job inputs.

• Refactor

  • UI copy and navigation shifted from "skills" to "workflows" across runner, settings, and scheduling.

• Tests

  • Added unit and end-to-end coverage for workflows, runner, scheduling, and orchestration.

[coderabbitai]

Actionable comments posted: 6

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (3)

src/openhuman/workflows/ops_create.rs (1)

47-89: ⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Reject unsupported inputs[].type values before writing workflow.toml.

The docs here say only string, integer, and boolean are supported, but the create/update path never enforces that. A bad RPC caller can persist an arbitrary kind and the runner then has no defined control to render on rediscovery.

Suggested fix

 fn validate_inputs(inputs: &mut [WorkflowCreateInputDef]) -> Result<(), String> {
     let mut seen: std::collections::HashSet<String> = std::collections::HashSet::new();
     for input in inputs.iter_mut() {
         let trimmed = input.name.trim();
         if trimmed.is_empty() {
             return Err("input name must not be empty".to_string());
         }
         if !seen.insert(trimmed.to_ascii_lowercase()) {
             return Err(format!("duplicate input name '{trimmed}'"));
         }
         let trimmed = trimmed.to_string();
         input.name = trimmed;
+        input.type_ = match input.type_.as_deref().map(str::trim).filter(|s| !s.is_empty()) {
+            Some("string" | "integer" | "boolean" as kind) => Some(kind.to_string()),
+            Some(other) => return Err(format!("unsupported input type '{other}'")),
+            None => None,
+        };
     }
     Ok(())
 }

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/openhuman/workflows/ops_create.rs` around lines 47 - 89, Reject and
validate unsupported input types before persisting workflow metadata: in the
create/update entrypoints (e.g., create_workflow, create_workflow_inner, and
workflows_update) iterate over CreateWorkflowParams.inputs (each
WorkflowCreateInputDef) and verify the type/kind is one of the allowed values
"string", "integer", or "boolean"; if any input has an unsupported type, return
a clear error (or reject the request) instead of writing
workflow.toml/skill.toml, ensuring the check runs both on create and update
paths so invalid kinds never get persisted.

src/core/event_bus/events_tests.rs (1)

214-242: ⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Keep the legacy event domain backward-compatible here.

These assertions lock the workflow lifecycle events onto "workflow". If any existing filters, dashboards, or trigger rules still key on the historic "skill" domain, this codifies a breaking change with no alias or migration path.

Based on learnings, "When reviewing changes to DomainEvent::variant_name() (or other “stable string” identifiers used by event bus consumers), treat those strings as a compatibility surface."

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/core/event_bus/events_tests.rs` around lines 214 - 242, The test changes
lock workflow lifecycle events to the new "workflow" domain and remove the
historic "skill" alias, breaking backward compatibility; update the
domain-returning logic (e.g., DomainEvent::variant_name / DomainEvent::domain or
wherever event domain string is produced) so that WorkflowLoaded,
WorkflowStopped, WorkflowStartFailed, and WorkflowExecuted continue to advertise
the legacy "skill" alias in addition to "workflow" (for example by returning
both names, emitting an alias field, or mapping "skill" to "workflow" on
input/output), and adjust tests to assert the presence of the legacy alias so
existing filters/dashboards/triggers keep working.

src/openhuman/workflows/ops_install.rs (1)

251-266: ⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Check all discovered user roots for slug collisions before writing.

This only guards ~/.openhuman/skills/<slug>, but discovery still loads ~/.openhuman/workflows/ and ~/.agents/skills/. If the slug already exists in one of those roots, this install succeeds into a shadowed path and the catalog can remain unchanged.

Collision guard sketch

-    let skills_root = home
-        .as_deref()
-        .ok_or_else(|| "write failed: unable to resolve home directory".to_string())?
-        .join(".openhuman")
-        .join("skills");
+    let home_dir = home
+        .as_deref()
+        .ok_or_else(|| "write failed: unable to resolve home directory".to_string())?;
+    for existing in [
+        home_dir.join(".openhuman").join("workflows").join(&slug),
+        home_dir.join(".openhuman").join("skills").join(&slug),
+        home_dir.join(".agents").join("skills").join(&slug),
+    ] {
+        if existing.exists() {
+            return Err(format!(
+                "workflow already installed as {slug:?} at {}",
+                existing.display()
+            ));
+        }
+    }
+    let skills_root = home_dir.join(".openhuman").join("skills");

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/openhuman/workflows/ops_install.rs` around lines 251 - 266, The current
install only checks the single target_dir (skills_root.join(&slug)) for
existence, but you must check all discovered user skill roots to prevent
shadowing; replace the single exists() guard with a scan over all user-scope
roots (e.g. the same roots used by discovery such as the computed skills_root
plus the ~/.openhuman/workflows and ~/.agents/skills locations or by calling the
existing discover_workflows/discovery helper) and if any root.join(&slug) exists
return an Err that names the colliding path; ensure you still compute the
intended target_dir for the actual install but abort early if any
discovered_root.join(&slug).exists() is true (use slug, skills_root, target_dir
and the discovery helper names to locate where to add the loop).

♻️ Duplicate comments (2)

app/src/components/intelligence/WorkflowsTab.tsx (2)

211-223: ⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Use workflow id (slug) for local uninstall reconciliation.

The onUninstalled callback filters the local list by result.name, but the uninstall flow is explicitly keyed by skill.id (the slug) because display name and slug can diverge. When they do, the removed card stays visible locally until refetch succeeds.

🔧 Proposed fix

       {uninstallCandidate && (
         <UninstallSkillConfirmDialog
           skill={uninstallCandidate}
           onClose={() => setUninstallCandidate(null)}
           onUninstalled={result => {
             log('uninstalled name=%s', result.name);
+            const removedId = uninstallCandidate.id;
             addToast({
               type: 'success',
               title: t('workflows.delete'),
               message: `"${result.name}" ${t('common.success')}`,
             });
-            setWorkflows(prev => prev.filter(s => s.id !== result.name));
+            setWorkflows(prev => prev.filter(s => s.id !== removedId));
             void refresh();
           }}
         />
       )}

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@app/src/components/intelligence/WorkflowsTab.tsx` around lines 211 - 223, The
local uninstall reconciliation currently filters workflows by result.name, which
can differ from the workflow slug used as the uninstall key; update the
onUninstalled handler in UninstallSkillConfirmDialog to remove the workflow
using the slug/id returned by the uninstall (use result.id or the property that
represents the workflow slug) — i.e., change the setWorkflows filter from s.id
!== result.name to s.id !== result.id (and keep logging/toast using result.name
if you want the display name intact).

---

51-62: ⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Error handling swallows backend failures.

The refresh() catch block returns [] and clears loading, causing the UI to show "no workflows installed" instead of surfacing the error. This hides outages and makes them appear as if the user has no workflows.

💡 Recommended fix

Add error state and render an error panel when backend calls fail:

+  const [loadError, setLoadError] = useState<string | null>(null);
+
   const refresh = useCallback(async (): Promise<SkillSummary[]> => {
     try {
       const list = await skillsApi.listSkills();
       log('listWorkflows ok count=%d', list.length);
+      setLoadError(null);
       setWorkflows(list);
       return list;
     } catch (err) {
-      log('listWorkflows error %s', err instanceof Error ? err.message : String(err));
+      const message = err instanceof Error ? err.message : String(err);
+      log('listWorkflows error %s', message);
+      setLoadError(message);
       return [];
     } finally {
       setLoading(false);
     }
   }, []);

Then render an error panel when loadError is set instead of the empty state.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@app/src/components/intelligence/WorkflowsTab.tsx` around lines 51 - 62, The
refresh() catch currently swallows backend failures by returning [] and only
clearing loading; add a component-level error state (e.g., loadError with setter
setLoadError) and in refresh() setLoadError(err) inside the catch (and clear it
on success), keep finally to setLoading(false), and update the component render
logic to show an error panel when loadError is set instead of rendering the
empty "no workflows" state; reference the refresh() function, setWorkflows,
setLoading and introduce setLoadError/loadError to locate the changes.

🧹 Nitpick comments (3)

app/src/pages/WorkflowsRun.tsx (1)

17-52: ⚡ Quick win

Consider adding unit test coverage for the new page component.

The WorkflowsRun page includes navigation logic (conditional back navigation with fallback) that would benefit from unit test coverage to verify the fallback behavior and ensure the page renders correctly. Per guidelines, new behavior in app/src/ should have unit tests via Vitest.

A test could verify:

• Renders with correct title and back button labels via t()
• Back button triggers navigate(-1) when history exists
• Back button navigates to /intelligence?tab=workflows when no history
• WorkflowRunnerBody is mounted

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@app/src/pages/WorkflowsRun.tsx` around lines 17 - 52, Add Vitest unit tests
for the WorkflowsRun component to exercise the conditional back navigation and
rendering: write tests that render WorkflowsRun (mocking useT to return
predictable labels and mocking WorkflowRunnerBody if needed), assert the title
and back button text from t(), simulate clicking the back button with
window.history.length > 1 and verify the mocked navigate was called with -1,
then simulate window.history.length === 1 and verify navigate was called with
'/intelligence?tab=workflows'; also include a test that WorkflowRunnerBody is
mounted. Mock useNavigate, useT, and optionally stub WorkflowRunnerBody to
isolate behavior.

src/openhuman/agent_registry/agents/orchestrator/prompt.md (1)

229-234: 💤 Low value

Add language specifier to fenced code block.

The code block contains JSON but lacks a language specifier, which affects syntax highlighting and linter compliance.

📝 Proposed fix

-```
+```json
 "images": [

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/openhuman/agent_registry/agents/orchestrator/prompt.md` around lines 229
- 234, The fenced code block containing the JSON "images" array needs a language
specifier for linters and highlighting; update the opening fence from ``` to
```json so the block that includes the "images" array and its objects (the
artifact/file entries and captions) is marked as JSON.

src/openhuman/workflows/bus.rs (1)

164-172: ⚡ Quick win

Rename the subscriber identity to the workflows domain.

TriggeredSkillSubscriber and name() still publish the old skills::triggered_skill label, so bus registration and grep-based tracing keep reporting the retired domain after the rename.

Rename sketch

-struct TriggeredSkillSubscriber {
+struct TriggeredWorkflowSubscriber {
     index: Arc<TriggeredWorkflowIndex>,
 }
 
 #[async_trait]
-impl EventHandler for TriggeredSkillSubscriber {
+impl EventHandler for TriggeredWorkflowSubscriber {
     fn name(&self) -> &str {
-        "skills::triggered_skill"
+        "workflows::triggered_workflow"
     }

As per coding guidelines, "Name the handler struct <Purpose>Subscriber and the name() return value "<domain>::<purpose>" for grep-friendly tracing output."

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/openhuman/workflows/bus.rs` around lines 164 - 172, The subscriber struct
and its name() still use the old skills domain; rename the struct
TriggeredSkillSubscriber to TriggeredWorkflowSubscriber and update its name() to
return "workflows::triggered_skill" (i.e., change the returned string from
"skills::triggered_skill" to "workflows::triggered_skill") so grep-friendly
tracing uses the workflows domain while preserving the purpose label.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@app/src/pages/WorkflowsRun.tsx`:
- Around line 33-35: The onClick handler in WorkflowsRun.tsx currently uses
window.history.length > 1 to decide between navigate(-1) and
navigate('/intelligence?tab=workflows'), which can still go back to an external
referrer; change the guard to check the in-app history index
(window.history.state?.idx > 0) like useSettingsNavigation does, and only call
navigate(-1) when that condition is true, otherwise navigate to
'/intelligence?tab=workflows' — update the onClick that references navigate(-1)
and window.history.length accordingly to use window.history.state?.idx > 0.

In `@src/openhuman/about_app/types.rs`:
- Around line 11-12: The Workflows enum variant currently has #[serde(rename =
"workflows")] but lacks a backward-compatible alias; add #[serde(alias =
"skills")] next to the rename on the Workflows variant (and repeat the same
alias change for the other affected variant referenced at the 71-71 location) so
serde will accept the legacy "skills" string during deserialization, and
update/add serde tests to assert that deserializing JSON with "skills" yields
the Workflows variant (and likewise for the other affected variant).

In `@src/openhuman/workflows/ops_create.rs`:
- Around line 281-317: When params.overwrite is true, don't call
render_workflow_md on parse failure (which overwrites the user's existing body);
instead, if super::ops_parse::parse_workflow_md(&workflow_md_path) and
super::ops_parse::parse_workflow_md(&legacy_md_path) both return None, read the
raw file contents from workflow_md_path (or legacy_md_path if workflow absent)
and return Some(raw_body) so the existing body is preserved and later combined
with render_workflow_frontmatter; update the preserved_body construction (the
branch assigning preserved_body) to implement this non-frontmatter fallback and
avoid calling render_workflow_md in that overwrite path.

In `@src/openhuman/workflows/ops_discover.rs`:
- Around line 223-233: The discovery currently uses Path::exists which follows
symlinks; update the checks for workflow_md, legacy_md, and legacy_manifest to
reject symlinks before loading: for each path (workflow_md, legacy_md,
legacy_manifest) call std::fs::symlink_metadata(...) and skip the path if
metadata.file_type().is_symlink() (or if symlink_metadata returns an error),
only then check existence/readability and call load_from_workflow_md (or
relevant loader). Reference symbols: workflow_md, legacy_md, legacy_manifest,
load_from_workflow_md, and use std::fs::symlink_metadata /
std::os::unix::fs::FileTypeExt/file_type().is_symlink() to implement the safe
check.

In `@src/openhuman/workflows/ops_install.rs`:
- Around line 408-421: The uninstall resolver only checks ~/.openhuman/workflows
and ~/.openhuman/skills so workflows discovered under ~/.agents/skills never
match and cannot be uninstalled; update the root-resolution logic in
ops_install.rs (the block that builds and .find(...)s roots for the trimmed id)
to include home.join(".agents").join("skills") alongside the existing openhuman
paths, so the same trimmed lookup used by discover_workflows_inner() will locate
workflows under ~/.agents/skills as well; keep the same existence check
(r.join(&trimmed).exists()) and error path unchanged.

In `@src/openhuman/workflows/run_log.rs`:
- Around line 24-66: Add debug-level logging to the cancellation registry: in
register_run_cancel(log the run_id and that a token was created/inserted and
whether it replaced an existing token), in unregister_run_cancel(log the run_id
and whether an entry was removed), and in cancel_run(log the run_id and whether
a token was found and cancelled or missed). Use the same logging crate used
elsewhere (debug! or trace!) and include the run_id and branch decisions
(created/inserted, replaced, removed, cancelled, missed) in the messages so
state transitions in register_run_cancel, unregister_run_cancel, and cancel_run
are observable without changing behavior.

---

Outside diff comments:
In `@src/core/event_bus/events_tests.rs`:
- Around line 214-242: The test changes lock workflow lifecycle events to the
new "workflow" domain and remove the historic "skill" alias, breaking backward
compatibility; update the domain-returning logic (e.g.,
DomainEvent::variant_name / DomainEvent::domain or wherever event domain string
is produced) so that WorkflowLoaded, WorkflowStopped, WorkflowStartFailed, and
WorkflowExecuted continue to advertise the legacy "skill" alias in addition to
"workflow" (for example by returning both names, emitting an alias field, or
mapping "skill" to "workflow" on input/output), and adjust tests to assert the
presence of the legacy alias so existing filters/dashboards/triggers keep
working.

In `@src/openhuman/workflows/ops_create.rs`:
- Around line 47-89: Reject and validate unsupported input types before
persisting workflow metadata: in the create/update entrypoints (e.g.,
create_workflow, create_workflow_inner, and workflows_update) iterate over
CreateWorkflowParams.inputs (each WorkflowCreateInputDef) and verify the
type/kind is one of the allowed values "string", "integer", or "boolean"; if any
input has an unsupported type, return a clear error (or reject the request)
instead of writing workflow.toml/skill.toml, ensuring the check runs both on
create and update paths so invalid kinds never get persisted.

In `@src/openhuman/workflows/ops_install.rs`:
- Around line 251-266: The current install only checks the single target_dir
(skills_root.join(&slug)) for existence, but you must check all discovered user
skill roots to prevent shadowing; replace the single exists() guard with a scan
over all user-scope roots (e.g. the same roots used by discovery such as the
computed skills_root plus the ~/.openhuman/workflows and ~/.agents/skills
locations or by calling the existing discover_workflows/discovery helper) and if
any root.join(&slug) exists return an Err that names the colliding path; ensure
you still compute the intended target_dir for the actual install but abort early
if any discovered_root.join(&slug).exists() is true (use slug, skills_root,
target_dir and the discovery helper names to locate where to add the loop).

---

Duplicate comments:
In `@app/src/components/intelligence/WorkflowsTab.tsx`:
- Around line 211-223: The local uninstall reconciliation currently filters
workflows by result.name, which can differ from the workflow slug used as the
uninstall key; update the onUninstalled handler in UninstallSkillConfirmDialog
to remove the workflow using the slug/id returned by the uninstall (use
result.id or the property that represents the workflow slug) — i.e., change the
setWorkflows filter from s.id !== result.name to s.id !== result.id (and keep
logging/toast using result.name if you want the display name intact).
- Around line 51-62: The refresh() catch currently swallows backend failures by
returning [] and only clearing loading; add a component-level error state (e.g.,
loadError with setter setLoadError) and in refresh() setLoadError(err) inside
the catch (and clear it on success), keep finally to setLoading(false), and
update the component render logic to show an error panel when loadError is set
instead of rendering the empty "no workflows" state; reference the refresh()
function, setWorkflows, setLoading and introduce setLoadError/loadError to
locate the changes.

---

Nitpick comments:
In `@app/src/pages/WorkflowsRun.tsx`:
- Around line 17-52: Add Vitest unit tests for the WorkflowsRun component to
exercise the conditional back navigation and rendering: write tests that render
WorkflowsRun (mocking useT to return predictable labels and mocking
WorkflowRunnerBody if needed), assert the title and back button text from t(),
simulate clicking the back button with window.history.length > 1 and verify the
mocked navigate was called with -1, then simulate window.history.length === 1
and verify navigate was called with '/intelligence?tab=workflows'; also include
a test that WorkflowRunnerBody is mounted. Mock useNavigate, useT, and
optionally stub WorkflowRunnerBody to isolate behavior.

In `@src/openhuman/agent_registry/agents/orchestrator/prompt.md`:
- Around line 229-234: The fenced code block containing the JSON "images" array
needs a language specifier for linters and highlighting; update the opening
fence from ``` to ```json so the block that includes the "images" array and its
objects (the artifact/file entries and captions) is marked as JSON.

In `@src/openhuman/workflows/bus.rs`:
- Around line 164-172: The subscriber struct and its name() still use the old
skills domain; rename the struct TriggeredSkillSubscriber to
TriggeredWorkflowSubscriber and update its name() to return
"workflows::triggered_skill" (i.e., change the returned string from
"skills::triggered_skill" to "workflows::triggered_skill") so grep-friendly
tracing uses the workflows domain while preserving the purpose label.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 26f02882-fbf4-4ece-a68e-f6badecd8269

📥 Commits

Reviewing files that changed from the base of the PR and between d000b5c and 0cfdcb2.

📒 Files selected for processing (82)

• app/src/components/intelligence/WorkflowsTab.tsx
• app/src/components/intelligence/__tests__/WorkflowsTab.test.tsx
• app/src/components/skills/CreateSkillModal.tsx
• app/src/components/skills/CreateWorkflowForm.tsx
• app/src/components/skills/SkillDetailDrawer.tsx
• app/src/components/skills/SkillResourcePreview.tsx
• app/src/components/skills/UninstallSkillConfirmDialog.tsx
• app/src/components/skills/WorkflowRunnerBody.tsx
• app/src/components/skills/__tests__/CreateSkillModal.test.tsx
• app/src/components/skills/__tests__/SkillDetailDrawer.test.tsx
• app/src/components/skills/__tests__/SkillResourcePreview.test.tsx
• app/src/components/skills/__tests__/UninstallSkillConfirmDialog.test.tsx
• app/src/components/skills/__tests__/WorkflowRunnerBody.test.tsx
• app/src/lib/i18n/ar.ts
• app/src/lib/i18n/bn.ts
• app/src/lib/i18n/de.ts
• app/src/lib/i18n/en.ts
• app/src/lib/i18n/es.ts
• app/src/lib/i18n/fr.ts
• app/src/lib/i18n/hi.ts
• app/src/lib/i18n/id.ts
• app/src/lib/i18n/it.ts
• app/src/lib/i18n/ko.ts
• app/src/lib/i18n/pl.ts
• app/src/lib/i18n/pt.ts
• app/src/lib/i18n/ru.ts
• app/src/lib/i18n/zh-CN.ts
• app/src/pages/Intelligence.tsx
• app/src/pages/WorkflowsRun.tsx
• app/src/services/api/skillsApi.ts
• src/core/event_bus/events.rs
• src/core/event_bus/events_tests.rs
• src/core/jsonrpc.rs
• src/openhuman/about_app/README.md
• src/openhuman/about_app/catalog_data.rs
• src/openhuman/about_app/catalog_tests.rs
• src/openhuman/about_app/types.rs
• src/openhuman/agent/harness/session/builder.rs
• src/openhuman/agent/harness/session/runtime.rs
• src/openhuman/agent/harness/session/turn.rs
• src/openhuman/agent/harness/session/types.rs
• src/openhuman/agent/harness/subagent_runner/ops.rs
• src/openhuman/agent/profiles.rs
• src/openhuman/agent/task_dispatcher.rs
• src/openhuman/agent/tools.rs
• src/openhuman/agent/tools/run_workflow.rs
• src/openhuman/agent_registry/agents/orchestrator/agent.toml
• src/openhuman/agent_registry/agents/orchestrator/prompt.md
• src/openhuman/channels/runtime/startup.rs
• src/openhuman/composio/tools/direct.rs
• src/openhuman/composio/tools/direct_tests.rs
• src/openhuman/inference/provider/factory.rs
• src/openhuman/todos/ops.rs
• src/openhuman/tools/impl/presentation/tests.rs
• src/openhuman/tools/ops.rs
• src/openhuman/tools/user_filter.rs
• src/openhuman/workflows/bus.rs
• src/openhuman/workflows/e2e_plumbing_tests.rs
• src/openhuman/workflows/e2e_run_tests.rs
• src/openhuman/workflows/inject.rs
• src/openhuman/workflows/ops.rs
• src/openhuman/workflows/ops_create.rs
• src/openhuman/workflows/ops_discover.rs
• src/openhuman/workflows/ops_install.rs
• src/openhuman/workflows/ops_parse.rs
• src/openhuman/workflows/ops_tests.rs
• src/openhuman/workflows/ops_types.rs
• src/openhuman/workflows/registry.rs
• src/openhuman/workflows/run_log.rs
• src/openhuman/workflows/schemas.rs
• src/openhuman/workflows/tools.rs
• src/openhuman/workspace/ops.rs
• tests/agent_harness_leftovers_raw_coverage_e2e.rs
• tests/agent_prompts_subagent_raw_coverage_e2e.rs
• tests/agent_round26_raw_coverage_e2e.rs
• tests/agent_session_round24_raw_coverage_e2e.rs
• tests/composio_raw_coverage_e2e.rs
• tests/inference_agent_raw_coverage_e2e.rs
• tests/json_rpc_e2e.rs
• tests/personality_e2e.rs
• tests/tools_approval_channels_raw_coverage_e2e.rs
• tests/tools_channels_raw_coverage_e2e.rs

💤 Files with no reviewable changes (4)

• app/src/components/skills/SkillResourcePreview.tsx
• app/src/components/skills/tests/SkillResourcePreview.test.tsx
• app/src/components/skills/SkillDetailDrawer.tsx
• app/src/components/skills/tests/SkillDetailDrawer.test.tsx

✅ Files skipped from review due to trivial changes (4)

• src/openhuman/about_app/README.md
• app/src/lib/i18n/fr.ts
• app/src/lib/i18n/ar.ts
• app/src/lib/i18n/it.ts

🚧 Files skipped from review as they are similar to previous changes (31)

• src/openhuman/tools/impl/presentation/tests.rs
• src/openhuman/agent/task_dispatcher.rs
• src/openhuman/composio/tools/direct.rs
• app/src/components/intelligence/tests/WorkflowsTab.test.tsx
• src/openhuman/agent/harness/session/runtime.rs
• src/openhuman/workflows/e2e_plumbing_tests.rs
• src/openhuman/channels/runtime/startup.rs
• app/src/lib/i18n/bn.ts
• src/openhuman/composio/tools/direct_tests.rs
• app/src/lib/i18n/zh-CN.ts
• src/openhuman/agent_registry/agents/orchestrator/agent.toml
• src/openhuman/agent/harness/session/types.rs
• src/openhuman/todos/ops.rs
• app/src/lib/i18n/id.ts
• app/src/lib/i18n/es.ts
• src/openhuman/agent/harness/session/builder.rs
• src/openhuman/inference/provider/factory.rs
• app/src/pages/Intelligence.tsx
• app/src/lib/i18n/pl.ts
• src/openhuman/agent/harness/subagent_runner/ops.rs
• src/openhuman/workflows/e2e_run_tests.rs
• app/src/lib/i18n/en.ts
• app/src/lib/i18n/pt.ts
• src/openhuman/agent/profiles.rs
• src/openhuman/agent/harness/session/turn.rs
• app/src/lib/i18n/ko.ts
• src/openhuman/agent/tools/run_workflow.rs
• app/src/lib/i18n/ru.ts
• src/openhuman/agent/tools.rs
• src/openhuman/workflows/ops_tests.rs
• src/openhuman/workflows/schemas.rs

[sanil-23]

Addressed CodeRabbit's review in 324d0203 — the actionable bugs in the new code:

Core

• ops_create: edit now refuses rather than overwriting an unparseable body with the scaffold (data-loss guard).
• ops_install: uninstall also resolves ~/.agents/skills/ (the legacy user root discovery surfaces).
• ops_discover: reject symlinked WORKFLOW.md/SKILL.md/skill.json before loading (symlink-follow disclosure path).
• run_log: read_terminal_outcome now requires the closing finished: line; added debug logs to the run-cancel registry.
• user_filter: kept skill_manage as a legacy alias for workflow_manage (persisted prefs back-compat).
• startup: wired the real triggered-workflow subscriber after discovery (was a no-op shim, so triggers: were ignored).

Frontend

• WorkflowsTab: listSkills() failures render an error/retry panel instead of the empty state; uninstall reconciles by id/slug not display name.
• Intelligence: ?tab= validated against the visible tab set so a forged ?tab=council can't open the dev-only tab in prod.

New tests: edit-refuses-unparseable-body, uninstall-resolves-~/.agents/skills, symlinked-manifest-rejected, terminal-outcome-requires-finished.

Deferred (per @senamakel on this migration PR): run_workflow spawn-backstop counting, event_bus variant_name() stability, and the about_app/ capability-catalog updates.

[sanil-23]

@coderabbitai review

Pushed fixes for the prior review (commit 324d020) plus an upstream re-merge (ca207d6). Addressed in-code: the edit body-clobber guard (ops_create), uninstall reaching ~/.agents/skills (ops_install), symlink-manifest rejection (ops_discover), terminal-outcome footer check + run-cancel debug logs (run_log), skill_manage legacy alias (user_filter), triggered-workflow subscriber wiring (startup), and the FE listSkills error panel / uninstall-by-id / tab-guard. Could you re-scan and resolve the threads that are now addressed? The agent_workflows variant_name/domain + about_app catalog items remain deferred per @senamakel for this migration PR.

[coderabbitai]

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# List all review comments (unresolved) on PR `#3324`
gh api graphql -f query='
{
  repository(owner: "tinyhumansai", name: "openhuman") {
    pullRequest(number: 3324) {
      reviewThreads(first: 50) {
        nodes {
          id
          isResolved
          comments(first: 1) {
            nodes {
              id
              databaseId
              body
              author { login }
              path
              line
            }
          }
        }
      }
    }
  }
}' | jq '.data.repository.pullRequest.reviewThreads.nodes[] | select(.isResolved == false) | {threadId: .id, path: .comments.nodes[0].path, line: .comments.nodes[0].line, author: .comments.nodes[0].author.login, body: .comments.nodes[0].body[:200]}'

Length of output: 1502

---

🏁 Script executed:

#!/bin/bash
# Get full body of both unresolved threads
gh api graphql -f query='
{
  repository(owner: "tinyhumansai", name: "openhuman") {
    pullRequest(number: 3324) {
      reviewThreads(first: 50) {
        nodes {
          id
          isResolved
          comments(first: 3) {
            nodes {
              databaseId
              body
              author { login }
            }
          }
        }
      }
    }
  }
}' | jq '.data.repository.pullRequest.reviewThreads.nodes[] | select(.isResolved == false)'

echo "---FILE CONTENT---"
# Check current WorkflowsRun.tsx for the navigation pattern
rg -n "history|navigate" app/src/pages/WorkflowsRun.tsx

Length of output: 5631

---

---

@sanil-23 — thanks for the detailed summary. I re-scanned both remaining open threads and here's the status:

---

🟡 Still open — fix not yet applied

app/src/pages/WorkflowsRun.tsx line 34 — back-navigation guard

The code still reads:

window.history.length > 1 ? navigate(-1) : navigate('/intelligence?tab=workflows')

The suggested fix (align with useSettingsNavigation's safer window.history.state?.idx > 0 guard) was not picked up in commit 324d0203. The original concern still applies — history.length > 1 fires even when the first entry is an external referrer, potentially bouncing the user out of the app. Quick one-liner fix:

window.history.state?.idx > 0 ? navigate(-1) : navigate('/intelligence?tab=workflows')

---

🔵 Intentionally deferred — keeping thread open as a reminder

src/openhuman/about_app/types.rs line 12 — #[serde(alias = "skills")] on the Workflows variant

Acknowledged as deferred per @senamakel for this migration PR (same scope as the event_bus variant_name() stability item). Thread stays open so it isn't forgotten in the follow-up.

---

Triggering a full re-review of the pushed commits now.

---

🧠 Learnings used

Learnt from: oxoxDev
Repo: tinyhumansai/openhuman PR: 833
File: app/src/pages/Skills.tsx:831-857
Timestamp: 2026-04-24T11:38:02.319Z
Learning: In `app/src/pages/Skills.tsx` and `app/src/components/skills/UninstallSkillConfirmDialog.tsx` (PR `#833`), the page-level integration test for the uninstall flow (`Skills.uninstall.test.tsx`) was intentionally deferred to a follow-up. The planned test should: mount the full Skills page using the `renderWithProviders` harness (same pattern as `Skills.composio-catalog.test.tsx`), click through the skill card → uninstall overflow menu → UninstallSkillConfirmDialog → confirm, and assert: `discoveredSkills` row is removed from the grid, detail drawer selection is cleared when it matched the removed skill, a success toast appears via ToastContainer, and `refreshDiscoveredSkills` is invoked. The blocker was the page harness setup time; the slug regression (id vs name) was already covered in `UninstallSkillConfirmDialog.test.tsx` commit b160a8ef.

---

Learnt from: senamakel
Repo: tinyhumansai/openhuman PR: 3324
File: src/core/event_bus/events.rs:1080-1080
Timestamp: 2026-06-04T05:02:47.951Z
Learning: In `tinyhumansai/openhuman` PR `#3324`, the `DomainEvent::variant_name()` rename in `src/core/event_bus/events.rs` from `SkillLoaded`/`SkillStopped`/`SkillStartFailed`/`SkillExecuted` to `WorkflowLoaded`/`WorkflowStopped`/`WorkflowStartFailed`/`WorkflowExecuted` was intentionally deferred as a compatibility follow-up. Future reviews should consider downstream persisted filters, dashboards, or trigger rules that may still key off the old stable `variant_name()` strings and may require an alias or migration path.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-02T22:01:29.371Z
Learning: Before opening AI-authored PRs from Codex web sessions or Linear-launched implementation agents, follow [`docs/agent-workflows/codex-pr-checklist.md`](docs/agent-workflows/codex-pr-checklist.md). This checklist is required for remote agents because OpenHuman has several merge gates: Prettier, Rust formatting, TypeScript typecheck, focused Vitest coverage, controller dispatch parity, and Tauri vendored dependency availability.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T03:23:47.354Z
Learning: Before opening AI-authored PRs from Codex web sessions or Linear-launched implementation agents, follow docs/agent-workflows/codex-pr-checklist.md. This checklist is required because OpenHuman has merge gates for Prettier, Rust formatting, TypeScript typecheck, focused Vitest coverage, controller dispatch parity, and Tauri vendored dependencies.

---

Learnt from: graycyrus
Repo: tinyhumansai/openhuman PR: 283
File: app/test/e2e/specs/skill-execution-flow.spec.ts:67-114
Timestamp: 2026-04-03T06:28:58.632Z
Learning: In `app/test/e2e/specs/skill-execution-flow.spec.ts`, the model→agent→tool→conversation E2E path (model-triggered tool_calls) is intentionally deferred to a future PR. It is tracked via an `it.skip` placeholder at the end of the spec. The blocker is that the mock backend does not yet support returning structured `tool_calls` in chat completion responses. The current spec only covers the deterministic skill runtime RPC surface: `skills_start` → `skills_list_tools` → `skills_call_tool` → `skills_stop`.

---

Learnt from: senamakel
Repo: tinyhumansai/openhuman PR: 1455
File: scripts/work/start.sh:150-154
Timestamp: 2026-05-10T03:14:32.586Z
Learning: In `scripts/work/start.sh` (repo: tinyhumansai/openhuman), `--agent codex` unconditionally invokes `codex exec --dangerously-bypass-approvals-and-sandbox`. This is intentional by design: the wrapper is a local contributor handoff tool and yolo/bypass mode was explicitly requested by the maintainer. The prompt already includes an instruction to treat the GitHub issue body as untrusted content. Do not flag this as a security issue in future reviews.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T21:18:08.576Z
Learning: Before opening AI-authored PRs from Codex web or Linear-launched agents, follow the checklist in `docs/agent-workflows/codex-pr-checklist.md` and report any blocked commands with exact error messages instead of claiming validation passed

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T21:18:08.576Z
Learning: Before opening AI-authored PRs from Codex web sessions or Linear-launched implementation agents, follow [`docs/agent-workflows/codex-pr-checklist.md`](docs/agent-workflows/codex-pr-checklist.md); this checklist is required because OpenHuman has several merge gates that are easy to miss in partial environments

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T04:47:45.876Z
Learning: Follow the Codex PR checklist (`docs/agent-workflows/codex-pr-checklist.md`) before opening AI-authored PRs; validate Prettier, Rust formatting, TypeScript typecheck, focused Vitest coverage, controller dispatch parity, and Tauri vendored dependency availability

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T21:50:55.026Z
Learning: Before opening AI-authored PRs from Codex web sessions or Linear-launched implementation agents, follow docs/agent-workflows/codex-pr-checklist.md

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T05:05:50.773Z
Learning: Before opening AI-authored PRs from Codex web sessions or Linear-launched implementation agents, follow the checklist in `docs/agent-workflows/codex-pr-checklist.md`

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T21:50:55.026Z
Learning: Before opening AI-authored PRs from Codex web sessions or Linear-launched implementation agents, follow `docs/agent-workflows/codex-pr-checklist.md`; the checklist is required for remote agents

✅ Action performed

Full review finished.

[sanil-23]

Thanks for the re-scan @coderabbitai.

✅ Applied (commit b45d317) — WorkflowsRun.tsx Back-nav now guards on the router history index instead of history.length:

(window.history.state?.idx ?? 0) > 0 ? navigate(-1) : navigate('/intelligence?tab=workflows')

so it won't bounce out of the app when the only prior entry is an external referrer (matches useSettingsNavigation's guard).

🔵 Deferred to a follow-up PR (out of scope for this migration, per @senamakel): the #[serde(alias = "skills")] back-compat on about_app::CapabilityCategory, the event_bus::variant_name()/domain() stability, the run_workflow spawn-backstop counting, and completing the internal skills→workflows identifier rename (the surviving PromptContext.skills field etc.). Keeping those threads open as reminders — they'll be picked up together in the rename PR.

[coderabbitai]

Tip

For best results, initiate chat on the files or code changes.

@sanil-23 — confirmed on both counts, thanks for the clean summary.

✅ WorkflowsRun.tsx back-nav — fix verified. Using window.history.state?.idx ?? 0 matches the useSettingsNavigation pattern and safely handles the external-referrer edge case. Resolving that thread.

🔵 Deferred follow-up items — noted. I'll record the new ones below so they aren't lost across PRs. Would you like me to open a GitHub issue to track them together?

🐇 ✨

---

✏️ Learnings added

Learnt from: sanil-23
URL: https://github.com/tinyhumansai/openhuman/pull/3324

Timestamp: 2026-06-04T15:15:38.156Z
Learning: In `tinyhumansai/openhuman` PR `#3324`, the `#[serde(alias = "skills")]` back-compat attribute on `CapabilityCategory::Workflows` in `src/openhuman/about_app/types.rs` was intentionally deferred to a follow-up rename PR (per senamakel). The alias ensures external callers that serialise `"skills"` continue to deserialise into `Workflows` after the rename; it should be added at the same time as the other `skills`→`workflows` identifier renames.

---

Learnt from: sanil-23
URL: https://github.com/tinyhumansai/openhuman/pull/3324

Timestamp: 2026-06-04T15:15:38.156Z
Learning: In `tinyhumansai/openhuman` PR `#3324`, the `run_workflow` spawn-backstop counter correctness (ensuring the atomic backstop counter is decremented on every code path, including early-return errors before the background task is spawned) was intentionally deferred to a follow-up PR per senamakel. Future reviews should check that `SPAWN_BACKSTOP_COUNT` is decremented in all error/early-exit branches in `src/openhuman/agent/tools/run_workflow.rs`.

---

Learnt from: sanil-23
URL: https://github.com/tinyhumansai/openhuman/pull/3324

Timestamp: 2026-06-04T15:15:38.156Z
Learning: In `tinyhumansai/openhuman` PR `#3324`, completing the internal `skills`→`workflows` identifier rename (e.g., the surviving `PromptContext.skills` field in `src/openhuman/agent/prompts/types.rs`, `AgentBuilder::skills()` in the session builder, and `Agent::skills()` accessor) was intentionally deferred to a dedicated rename follow-up PR per senamakel. These symbols keep `skills` in their Rust names while holding `Workflow` values; the rename should be done in one sweep to minimise churn.

Note: Learnings are effective only in the context of similar code segments. To apply general coding guidelines, please configure review instructions. You can manage existing learnings in the app.

---

🧠 Learnings used

Learnt from: senamakel
Repo: tinyhumansai/openhuman PR: 3324
File: src/core/event_bus/events.rs:1080-1080
Timestamp: 2026-06-04T05:02:47.951Z
Learning: In `tinyhumansai/openhuman` PR `#3324`, the `DomainEvent::variant_name()` rename in `src/core/event_bus/events.rs` from `SkillLoaded`/`SkillStopped`/`SkillStartFailed`/`SkillExecuted` to `WorkflowLoaded`/`WorkflowStopped`/`WorkflowStartFailed`/`WorkflowExecuted` was intentionally deferred as a compatibility follow-up. Future reviews should consider downstream persisted filters, dashboards, or trigger rules that may still key off the old stable `variant_name()` strings and may require an alias or migration path.

---

Learnt from: oxoxDev
Repo: tinyhumansai/openhuman PR: 833
File: app/src/pages/Skills.tsx:831-857
Timestamp: 2026-04-24T11:38:02.319Z
Learning: In `app/src/pages/Skills.tsx` and `app/src/components/skills/UninstallSkillConfirmDialog.tsx` (PR `#833`), the page-level integration test for the uninstall flow (`Skills.uninstall.test.tsx`) was intentionally deferred to a follow-up. The planned test should: mount the full Skills page using the `renderWithProviders` harness (same pattern as `Skills.composio-catalog.test.tsx`), click through the skill card → uninstall overflow menu → UninstallSkillConfirmDialog → confirm, and assert: `discoveredSkills` row is removed from the grid, detail drawer selection is cleared when it matched the removed skill, a success toast appears via ToastContainer, and `refreshDiscoveredSkills` is invoked. The blocker was the page harness setup time; the slug regression (id vs name) was already covered in `UninstallSkillConfirmDialog.test.tsx` commit b160a8ef.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-02T22:01:29.371Z
Learning: Before opening AI-authored PRs from Codex web sessions or Linear-launched implementation agents, follow [`docs/agent-workflows/codex-pr-checklist.md`](docs/agent-workflows/codex-pr-checklist.md). This checklist is required for remote agents because OpenHuman has several merge gates: Prettier, Rust formatting, TypeScript typecheck, focused Vitest coverage, controller dispatch parity, and Tauri vendored dependency availability.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T03:23:47.354Z
Learning: Before opening AI-authored PRs from Codex web sessions or Linear-launched implementation agents, follow docs/agent-workflows/codex-pr-checklist.md. This checklist is required because OpenHuman has merge gates for Prettier, Rust formatting, TypeScript typecheck, focused Vitest coverage, controller dispatch parity, and Tauri vendored dependencies.

---

Learnt from: graycyrus
Repo: tinyhumansai/openhuman PR: 2703
File: app/src/components/settings/panels/DevWorkflowPanel.tsx:55-71
Timestamp: 2026-05-27T14:39:40.859Z
Learning: In `app/src/components/settings/panels/DevWorkflowPanel.tsx`, the `DevWorkflowConfig` is intentionally persisted via ad-hoc `localStorage` (key `openhuman:dev-workflow-config`) as a Phase 1 temporary stopgap. Phase 2 (tracked in PR `#2707`) will migrate this config to the Rust core via the `skills_run` RPC. A Redux Toolkit slice was deliberately skipped to avoid churn on a short-lived intermediate layer. Do not flag this localStorage usage as a violation of the Redux-persist guideline.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T04:47:45.876Z
Learning: Follow the Codex PR checklist (`docs/agent-workflows/codex-pr-checklist.md`) before opening AI-authored PRs; validate Prettier, Rust formatting, TypeScript typecheck, focused Vitest coverage, controller dispatch parity, and Tauri vendored dependency availability

---

Learnt from: graycyrus
Repo: tinyhumansai/openhuman PR: 283
File: app/test/e2e/specs/skill-execution-flow.spec.ts:67-114
Timestamp: 2026-04-03T06:28:58.632Z
Learning: In `app/test/e2e/specs/skill-execution-flow.spec.ts`, the model→agent→tool→conversation E2E path (model-triggered tool_calls) is intentionally deferred to a future PR. It is tracked via an `it.skip` placeholder at the end of the spec. The blocker is that the mock backend does not yet support returning structured `tool_calls` in chat completion responses. The current spec only covers the deterministic skill runtime RPC surface: `skills_start` → `skills_list_tools` → `skills_call_tool` → `skills_stop`.

---

Learnt from: senamakel
Repo: tinyhumansai/openhuman PR: 2112
File: app/src-tauri/permissions/allow-core-process.toml:103-122
Timestamp: 2026-05-20T02:56:09.546Z
Learning: In the `tinyhumansai/openhuman` repository, splitting `logs_folder_path`/`reveal_logs_folder` and `meet_call_open_window`/`meet_call_close_window` out of the `allow-core-process` permission bundle (`app/src-tauri/permissions/allow-core-process.toml`) into dedicated narrower permission identifiers (e.g., `allow-logs-access` and `allow-meet-window-control`) is a known follow-up task (deferred from PR `#2112`). It requires: (1) creating new permission TOML files, (2) updating capability grants in `app/src-tauri/capabilities/` to bind them only to the UI surfaces that need them, and (3) verifying no `allow-core-process` consumer regresses.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T21:18:08.576Z
Learning: Before opening AI-authored PRs from Codex web or Linear-launched agents, follow the checklist in `docs/agent-workflows/codex-pr-checklist.md` and report any blocked commands with exact error messages instead of claiming validation passed

---

Learnt from: sanil-23
Repo: tinyhumansai/openhuman PR: 1825
File: src/openhuman/composio/tools.rs:1028-1039
Timestamp: 2026-05-15T22:00:15.051Z
Learning: In `src/openhuman/composio/tools.rs` (PR `#1825`), the stale `Arc<Config>` snapshot concern was resolved in commit `18ced400` (Wave 4). Each tool's `execute()` method now calls `config_rpc::load_config_with_timeout()` at the top of every invocation to reload the live on-disk config before calling `create_composio_client()`. The `Arc<Config>` held by `self` (captured at agent-init via `Arc::new(config.clone())`) is no longer used for mode selection or key lookup — it is purely an init-time reference. The same per-call reload pattern was applied to `ProviderContext::execute()` and `ProviderContext::backend_client()` in `src/openhuman/composio/providers/types.rs`. 13 tests were migrated to the `TEST_ENV_LOCK` + `OPENHUMAN_WORKSPACE`-tempdir + `config.save()` isolation pattern to ensure `load_config_with_timeout()` reads a controlled snapshot in tests.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T03:23:47.354Z
Learning: Applies to src/openhuman/about_app/**/*.rs : Update `src/openhuman/about_app/` (capability catalog) when a change adds, removes, renames, relocates, or materially changes a user-facing feature.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T03:23:47.354Z
Learning: Applies to src/openhuman/about_app/**/*.rs : Update `src/openhuman/about_app/` when a change adds, removes, renames, relocates, or materially changes a user-facing feature so the runtime capability catalog remains the source of truth

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T03:23:47.354Z
Learning: Applies to src/openhuman/about_app/**/*.rs : Update src/openhuman/about_app/ capability catalog when a change adds, removes, renames, relocates, or materially changes a user-facing feature.

---

Learnt from: CR
Repo: tinyhumansai/openhuman PR: 0
File: AGENTS.md:0-0
Timestamp: 2026-06-03T03:23:47.354Z
Learning: Applies to src/openhuman/about_app/**/*.rs : When a change adds, removes, renames, relocates, or materially changes a user-facing feature, update `src/openhuman/about_app/` in the same work so the runtime capability catalog remains the source of truth