docs: fix 30 documentation inaccuracies found by source code audit

[aidandaly24]

Description

Comprehensive audit of all .md documentation files against actual source code (Zod schemas, CLI commands, CDK constructs). Fixes 50+ inaccuracies across 14 files where docs had drifted from the implementation, including alignment with several upstream refactors that landed during this effort.

Upstream refactors accounted for

PR	Change	Doc Impact
#709	Remove type fields from resource schemas, rename credential type → authorizerType	Removed stale type fields from all JSON examples; renamed credential discriminator
#705	add identity → add credential, --force → -y/--yes, --plan → --dry-run	Updated all command references and flag tables
#706	agents → runtimes array, --agent → --runtime flags	Updated all flag tables, examples, and JSON config across 7 doc files
#707	Dev command --invoke → positional [prompt] arg	Already reflected in docs
#713	Revert CUSTOM strategy and semanticOverride	No doc changes needed (CUSTOM was never documented)

New features documented (merged after initial audit)

PR	Feature	Doc Files
#763, #780	Import subcommands (import runtime, import memory, import evaluator, import online-eval)	commands.md
#750	Exec mode for invoke and dev (--exec, --timeout)	commands.md, local-development.md
#739	Code-based evaluator support (--type, --lambda-arn, --timeout)	commands.md, evals.md
#732	Ground truth reference inputs (--assertion, --expected-trajectory, --expected-response)	commands.md, evals.md
#531, #534	Memory record streaming (--data-stream-arn, --stream-content-level, --delivery-type)	commands.md, memory.md
#782	--skip-install flag on create	commands.md
#774	--identity-name option on fetch access	commands.md

Key fixes by category

Schema/config drift (configuration.md, gateway.md, container-builds.md):

• Added missing EPISODIC memory strategy and reflectionNamespaces field
• Added lifecycleConfiguration (idleRuntimeSessionTimeout, maxLifetime) to agent spec
• Fixed CUSTOM_JWT fields: allowedAudience/allowedClients/allowedScopes/customClaims are conditionally required
• Added missing customClaims field to CUSTOM_JWT config
• Fixed MCP Server outbound auth: removed api-key (only oauth/none valid per schema)
• Added enableSemanticSearch, exceptionLevel, policyEngineConfiguration gateway fields
• Fixed memory strategies default: [] not @min 1 (empty array valid for short-term memory)
• Removed stale "type": "AgentCoreRuntime" / "AgentCoreMemory" / "CustomEvaluator" / "OnlineEvaluationConfig" from all JSON examples

Flag renames (commands.md, evals.md, local-development.md):

• --agent → --runtime, -a → -r across all commands (status, dev, invoke, logs, traces, run eval, evals history, package, add online-eval, logs evals)
• --agent-arn → --runtime-arn, --agent-runtime-id → --runtime-id
• --plan → --dry-run, --force → -y/--yes
• add identity → add credential, remove identity → remove credential
• --agent-client-id → --client-id, --agent-client-secret → --client-secret (gateway CUSTOM_JWT)
• --exception-level valid values: NONE/DEBUG (not ALL)
• --lines → --limit in logs evals
• Note: --agent-id and --agent-alias-id are unchanged (Bedrock Agent import flags)

Missing flags added (commands.md):

• add agent: --authorizer-type, --discovery-url, --allowed-audience, --allowed-clients, --allowed-scopes, --custom-claims, --client-id, --client-secret, --idle-timeout, --max-lifetime
• create: --idle-timeout, --max-lifetime, --skip-install
• add gateway: --runtimes, --custom-claims, --policy-engine, --policy-engine-mode
• dev and invoke: -H, --header, --exec
• invoke: --bearer-token, --exec, --timeout
• add evaluator: --type, --lambda-arn, --timeout (code-based evaluator support)
• run eval: --assertion, --expected-trajectory, --expected-response (ground truth inputs)
• add memory: --data-stream-arn, --stream-content-level, --delivery-type, --stream-delivery-resources
• fetch access: --identity-name

New sections added:

• commands.md: Full import command section with 4 subcommands (runtime, memory, evaluator, online-eval)
• evals.md: Evaluator Types section (LLM-as-a-Judge vs code-based), Ground Truth Reference Inputs subsection
• memory.md: Memory Record Streaming section with CLI flags and JSON config example
• local-development.md: Exec Mode section covering dev exec, invoke exec, and TUI shortcuts

Stale values (commands.md, evals.md, frameworks.md):

• Fixed model ID claude-3-5-sonnet-20250929 → claude-3-5-sonnet-20250514
• Added policy-engine/policy to status --type enum
• Removed ghost --tool-filter-description flag (not in schema or CLI)
• Fixed evaluator level values: added TOOL_CALL, removed SPAN

Structural/naming (AGENTS.md, README.md, assets):

• Removed AutoGen from supported frameworks list (5 frameworks, not 6)
• Fixed "agents" → "runtimes" in gateway JSON example
• Renamed identity → credential in all command references
• Fixed file paths and directory references
• Fixed regex constraints and enum values to match Zod schemas

Files modified

File	Changes
docs/configuration.md	Schema field tables, JSON examples, type removals, credential rename
docs/commands.md	All flag tables and examples, import subcommands, exec mode, streaming, evaluator types, ground truth
docs/evals.md	Flag tables, examples, evaluator types section, ground truth inputs, --runtime rename
docs/gateway.md	identity → credential, agents → runtimes in JSON example
docs/local-development.md	--runtime flag in dev examples, exec mode section
docs/memory.md	Removed stale types, added memory record streaming section
docs/container-builds.md	Removed stale "type": "AgentCoreRuntime" from JSON example
docs/frameworks.md	Model ID fix, framework list correction
AGENTS.md	identity → credential in primitives
README.md	identity → credential
src/assets/README.md	identity → credential in vended template
src/assets/__tests__/__snapshots__/assets.snapshot.test.ts.snap	Updated snapshot

Related Issue

N/A — discovered via systematic source code audit

Documentation PR

N/A — this PR is the documentation fix

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Other (please describe):

Testing

• I ran npm run test:unit and npm run test:integ
• I ran npm run typecheck
• I ran npm run lint
• If I modified src/assets/, I ran npm run test:update-snapshots and committed the updated snapshots

All changes are documentation-only (.md files + snapshot update). Each fix was verified against the corresponding Zod schema definition or CLI source code (src/cli/commands/*/command.tsx, src/cli/primitives/*.ts).

Note: 19 pre-existing test failures exist on upstream/main from the schema refactor (#709) — not introduced by this PR. The 2 e2e failures are expected on fork PRs (no AWS credentials).

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the
terms of your choice.

[github-actions]

Package Tarball

aws-agentcore-0.7.1.tgz

How to install

npm install https://github.com/aws/agentcore-cli/releases/download/pr-697-tarball/aws-agentcore-0.7.1.tgz

[Hweinstock]

lgtm! Its possible we'll need another update after the nits are addressed.