fix: third-wave findings (example + canvas)

[ScriptedAlchemy]

Summary

• fix(examples): stage audiobook evidence work on disk, not os.tmpdir #353 P1: acoustic-identify forwards the receipt path into every per-candidate sampleMatch, so staging follows the receipt's disk instead of falling back beside source media; regression pins the staging directory
• Latent defect surfaced by that regression and fixed: identify receipts with titleless/asin-less candidates wrote undefined-valued keys that the strict JSON receipt writer rejects — absent fields are now omitted (identified + attempts)
• docs(canvas): correct four walkthrough accuracy findings #356: canvas scopes skills honestly (Markdown documents unless rendered SKILL.tsx) and corrects the tool/before note — the wrapper always answers with an explicit permissionDecision, silence is reserved for observation-only families; managed canvas copy synced

Test plan

• evidence-parity 8/8 incl. new receipt-staging regression (red-without-fix)
• full example check, typecheck, lint

[changeset-bot]

⚠️ No Changeset found

Latest commit: 5e406d3

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-03T04:38:55.156722Z	5e406d3	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5e406d3a84

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Limit the silence claim to this tool/before step

The last sentence is broader than the implementation: projectEventDocument returns undefined for decision-capable stop when the outcome is not deny (packages/agent-bundle/src/events/projection.ts:398-402), and for prompt/submit when neither a block reason nor context is emitted (projection.ts:472-474). These families are not observation-only, so readers could incorrectly infer that every actionable hook always emits a response. Keep the explicit-allow statement scoped to Claude/Codex tool/before without claiming silence is exclusive to observation-only families.

Useful? React with 👍 / 👎.

[ScriptedAlchemy]

Fixed in #397 (merged as d25a9c6). The walkthrough canvas wording now matches events/projection.ts: silence is the default for observation-only families and is also valid for decision-capable families (stop, prompt/submit) that choose not to decide. The same edit was applied to the managed copy at ~/.cursor/projects/fast-projects-agent-bundle/canvases/agent-bundle-walkthrough.canvas.tsx; cmp confirms the two files are byte-identical.