Repository navigation
testing: one named tool per job for browser and profiling #7580
Description
Parent
Refs #7555. Source: PLAN Phase 13 (first item), topic poteto-review. Thread: Tool per job (browser
and profiling).
Depends on
- config-cascade: settings resolution and convention home #7571 Settings resolution and convention home
- Child debugging: probe disallowed-tools MCP pattern and large-heap converter #7593 (Q561 probe) gates this thread's setting slice.
Outcome
Each job names one tool through a setting with a named default (browser_driver: playwright-cli):
skills check the tool resolves and stop if it is missing, and heap and trace analysis run in
analyze-profile's SQLite converter, with no MCP handoff.
Decisions
- One named tool per job, no MCP handoff: heap compare and trace analysis in analyze-profile's SQLite
converter;playwright-cliwith CDP sessions for driving and capture; skills name the command,
check it resolves, stop if missing, and say other browser tools are not substitutes (Q561).
Origin: Lauren's (Playwright plus CDP), owner's ("follow Lauren or better"). - Tool per job is a setting with a named default (e.g.
browser_driver: playwright-cli), read
through the adapter; when the default is missing, use the tool the user has, pick one per run and
say which, never rotate mid-run (Q569). Setting scope: repo. - Rejected: Chrome DevTools MCP as a handoff (its artifacts still accepted as input); Claude in
Chrome; Playwright MCP; Puppeteer CDPSession or chrome-remote-interface; memlab and agent-browser
until a run shows the converter cannot answer. - Research tags: skill
disallowed-toolswith an MCP server-wide pattern; the converter on a large
heap; Cursor support fordisallowed-tools(assumption 6, Q579). - This thread closes with the app-control thread's first slice (tool-per-job-setting).
Key interfaces
Variation verdicts: "Browser driver and profiling tools: abstract now; tool-per-job setting read
through the adapter; one named tool, no MCP handoff (Q561, Q569)." Settings: "Tool per job (e.g.
browser_driver: playwright-cli), repo."
Done when
- Child debugging: probe disallowed-tools MCP pattern and large-heap converter #7593 carries a
Probe result:comment for both parts. - The tool-per-job setting resolves through the config cascade with default
browser_driver: playwright-cli, and a skill that reads it stops with a named message when the
tool is missing (a test shows both). - Every slice sub-issue is closed by a merged PR.
Out of scope
- The app-control plugin and capture move (app-control thread).
Activity
github-actions commentedon Oct 16, 2026
Proposed work class: work-class: scoped
Basis: automated intake triage; the lanes App applied this class.
Size: Medium
Complexity: Medium
Uncertainty: Knowable
Flags: none
Blocked by: #7571, #7593
I understand this as adding a per-job tool setting (default
browser_driver: playwright-cli) that resolves through the config cascade. Skills that read it would stop with a named message when the tool is missing. Heap and trace analysis would stay in analyze-profile's SQLite converter, with no MCP handoff. The decisions and done-when items are specific, so I found no open intent question. The work depends on #7571 (settings resolution) and on the #7593 probe result. I found no likely duplicates in the open issues I was given.Automated triage, applied by the lanes App.
feat(testing): resolve a browser_driver setting and stop by name when…
kyle-sexton commentedon Oct 24, 2026
Released for re-triage under the owner's 2026-10-24 rule: an issue stops for a person only for an intent or vision question, a security call, a host or attended run, or spend outside prepaid credit. The hold label came off and intake triage runs again now; it re-applies a hold, with a numbered question, when one of those applies.
To hold this issue again, add
needs-human(orstatus: needs-decision) and say why in a comment.melodic-automation-lanes commentedon Oct 24, 2026
Not admitted to the issue lane yet: blocker #7571 is not closed by a listed human or a merge; sub-issue #8651 is open.
melodic-automation-lanes commentedon Oct 24, 2026
The latest triage verdict holds nothing: the issue is at
status: ready.kyle-sexton commentedon Oct 26, 2026
Released for re-triage under the owner's 2026-10-24 rule: an issue stops for a person only for an intent or vision question, a security call, a host or attended run, or spend outside prepaid credit. The hold label came off and intake triage runs again now; it re-applies a hold, with a numbered question, when one of those applies.
To hold this issue again, add
needs-human(orstatus: needs-decision) and say why in a comment.