The redaction and privacy cases grew session-activity-tracker.test.js to 845 lines, past the 800 that AGENTS.md allows. They move to session-activity-tracker-output.test.js; the original file is back to 758 lines.
The tracker serialized any structured result with JSON.stringify before
redaction, so an encoding nested in it, such as a double-escaped JSON
token, reached tool-usage.jsonl whole, and Claude Code's Bash stdout went
in as text. Keep text only from the adapter's tool_output, as main did,
and from Claude Code's tool_response only the sizes of stdout and stderr
and its interrupted and image flags. Any other structure is left out.
Restore the shipped payload-first contract from #3464/#3495 and the explicit source #3755 scope clarification. Two provider conversations must retain distinct IDs despite an inherited registered ECC parent ID. Normal and ambient producer24 tests, canonical resolver10 and independent differential review pass. Keep native importer guards; explicit provider/native correlation is queued separately.
Verify review discussion4419348312 against actual native producer/importer behavior. Prefer the independently sanitized ECC_SESSION_ID for the ECC-owned metrics feed, retain payload then legacy fallback for standalone hooks, and leave sibling hooks unchanged.
Actual Node producer plus read-only cached native binary reproduces zero imported metrics before the fix and exactly one tool/file/log after repeated sessions/status commands. Four producer regression controls and normal/adversarial inherited-env suites pass22/22. Focused coverage96.47percent lines and91.47percent branches; parent independent source review passed.
The bounded name parts (64 characters around the credential word) let a longer name such as API_KEY_ plus 65 characters go unredacted, and the field-bound bare value left `,beta` of CLIENT_SECRET=alpha,beta in the summary.
Find each name and separator first, starting a name only where a run of name characters starts, so every run is scanned once whatever its length, and test the whole name for the credential word. A bare value after a quoted JSON key still ends at the next field; any other runs to the next space.
A raw Docker configuration, as `cat ~/.docker/config.json` prints it, keeps its credential under `"auth":`, which only the auth_config name matched. Count `auth` as a credential word when it ends the word, so `"auth":`, `AUTH=` and `oauth=` are redacted while `author` and `auth_method` are not.
Three review findings on the generic credential rule:
- The unbounded name parts around the credential word made the expression rescan the rest of the text after each dot or hyphen. 128,000 characters of `a.` took about 8 seconds, and 256,000 ran past the hook timeout. The name parts are now bounded to 64 characters each, which brings the same input down to about 30 ms.
- A bare value ran to the next whitespace, so `{"token":null,"status":"failed","exit_code":1}` lost its status and exit code. A bare value now ends at the next comma, brace, bracket, semicolon or quote. A JSON object or array value is still redacted to the end.
- A quoted value stopped at an escaped quote, so `{"api_key":"part1\\" part2"}` kept `part2`. Quoted values now skip escaped characters.
An unquoted value ran to the next whitespace, so `DOCKER_AUTH_CONFIG={"auths": {"registry": {"auth": "..."}}}` kept everything after its first space, the auth value included. A value that opens with { or [ is now redacted to the end of the summary.
The generic redaction matched a credential word only right before the separator, so `AWS_SECRET_ACCESS_KEY=...` or `DOCKER_AUTH_CONFIG=...` kept its value in the activity summary. Match a name that contains secret, token, password, api_key, access_key, private_key, credential or auth_config anywhere, and take an unquoted value up to the next whitespace. `max_tokens`, `tokenizer` and `author` stay as they are.
With Claude Code's tool_response read, command output such as `cat .env` now reaches the activity log's output_summary. Its redaction covered a few token forms, not the ones governance-capture detects: generic key, secret and token assignments (`api_key = "..."`, `OPENAI_API_KEY=...`, `"token": "..."`), JWTs, private keys, and ghu_/ghr_ tokens. Redact those too.
Claude Code passes a PostToolUse hook the result of its tool as tool_response. For Bash, that is an object with stdout, stderr, interrupted and isImage. Three hooks read only tool_output, the shape that the Cursor adapter builds, so under Claude Code they never saw any output:
- post-bash-pr-created never printed the PR URL and the review command after `gh pr create`;
- governance-capture never ran its secret detection on command output, such as that of `cat .env`;
- session-activity-tracker logged an empty output_summary for every tool.
Add getToolResult and getToolResultText to hook-input.js. They read tool_response first, then tool_output. Use them in the three hooks. governance-capture now also reads the adapter's { output }, where it used to accept only a string.
session-activity-tracker.js is correctly registered in
posttooluse-dispatcher.js's SYNC_HOOKS and fires on every tool call --
but buildActivityRow() only ever read the session id from
ECC_SESSION_ID/CLAUDE_SESSION_ID environment variables, which Claude
Code does not actually set for hook subprocesses. Every row silently
came back null and nothing was ever written to
~/.claude/metrics/tool-usage.jsonl, with no error anywhere.
The session id is available in the hook's own stdin JSON payload
(input.session_id) in real invocations. This prefers that, keeping the
env vars as a fallback. New test added; all 18 existing + new tests
pass.
Also adds a live Activity view to the capabilities dashboard
(scripts/dashboard-web.js) reading that same metrics file, so Skill
vs. Agent vs. raw-tool usage is visible in real time instead of
requiring a manual tail+jq -- verified live against a real session
before this commit. Documents the loopback-only, no-auth design and
the standard SSH-port-forwarding pattern for viewing it from another
device you control.