From 0325c97371ed029c797058f9d12218e978d872af Mon Sep 17 00:00:00 2001 From: Claude Date: Sat, 15 Aug 2026 07:34:07 +0000 Subject: [PATCH 1/2] docs: open interview ledger for repo-wide audit-noise run Round 1 asked (Q1-Q6): deliverable shape, corpus scope, report destination, orchestration parameters, skill fallback update scope, rate-limit telemetry issue. Awaiting answers before the orchestrated sweep runs. Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_01H9A7ewGsRVk4KWVUhUcgpq --- .../interview-checklist.md | 35 +++++++++++++++++++ 1 file changed, 35 insertions(+) create mode 100644 .work/docs-hygiene-audit-noise/interview-checklist.md diff --git a/.work/docs-hygiene-audit-noise/interview-checklist.md b/.work/docs-hygiene-audit-noise/interview-checklist.md new file mode 100644 index 000000000..8f7883fc7 --- /dev/null +++ b/.work/docs-hygiene-audit-noise/interview-checklist.md @@ -0,0 +1,35 @@ +# /planning:interview Checklist — docs-hygiene-audit-noise + +Topic: repo-wide `/docs-hygiene:audit-noise` orchestrated run + skill fallback update. + +## Steps + +- [x] Step 1: Survey before you ask — repo has 1131 tracked `.md` (1027 after excluding evals fixtures + changelogs); tree clean; skill's literal default is a no-op on clean tree; detect.sh serial run over full corpus exceeds 5 min (needs chunking); `/session-flow:orchestrate` armed; user clarified intent mid-turn (orchestrated full-repo run + skill fallback design) +- [ ] Step 1.5: Auto-detect — user asked to be interviewed; forced Q&A (`me`-leaning) +- [ ] Step 2: Drive the frontier-rounds loop +- [ ] Step 3: Recognize the stop condition +- [ ] Step 4: Persist the contract +- [ ] Step 5: Hand off + +## Open-question register + +- Q1 | open | round 1 | Deliverable: report-only, or report + apply Tier 1 treatments in this session? | +- Q2 | open | round 1 | Corpus scope/exclusions for the repo-wide run? | +- Q3 | open | round 1 | Where does the audit report land (committed memory slice vs PR-only vs artifact)? | +- Q4 | open | round 1 | Orchestration shape: concurrency cap, model tier, verification pass? | +- Q5 | open | round 1 | Skill update scope: audit-noise fallback only, or propagate to sibling audit skills? | +- Q6 | open | round 1 | File the rate-limit-telemetry enhancement issue now, or defer? | + +## Decision tree (`me` mode) + +- [ ] Deliverable shape (Q1) +- [ ] Corpus scope (Q2) +- [ ] Report destination (Q3) +- [ ] Orchestration parameters (Q4) +- [ ] Skill fallback update scope (Q5) +- [ ] Rate-limit telemetry issue (Q6) + +## Session-shorthand glossary + +- "fallback" — the behavior `/docs-hygiene:audit-noise` should adopt when invoked with no target AND a clean tree (today: friendly no-op) +- "slices" — topic-docs working directories: `docs/topics//` (contract tier) and `.work//` (memory tier) From f0b8d07002876f1d05b0bc1023d7575448b49d53 Mon Sep 17 00:00:00 2001 From: Claude Date: Sat, 15 Aug 2026 07:37:10 +0000 Subject: [PATCH 2/2] fix: untrack memory-tier interview ledger per topic-docs convention The memory tier (.work//) is never committed and self-ignoring (docs/conventions/topic-docs/README.md tier table); the ledger commit violated that. The slice now carries its self-ignore and stays local. Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_01H9A7ewGsRVk4KWVUhUcgpq --- .../interview-checklist.md | 35 ------------------- 1 file changed, 35 deletions(-) delete mode 100644 .work/docs-hygiene-audit-noise/interview-checklist.md diff --git a/.work/docs-hygiene-audit-noise/interview-checklist.md b/.work/docs-hygiene-audit-noise/interview-checklist.md deleted file mode 100644 index 8f7883fc7..000000000 --- a/.work/docs-hygiene-audit-noise/interview-checklist.md +++ /dev/null @@ -1,35 +0,0 @@ -# /planning:interview Checklist — docs-hygiene-audit-noise - -Topic: repo-wide `/docs-hygiene:audit-noise` orchestrated run + skill fallback update. - -## Steps - -- [x] Step 1: Survey before you ask — repo has 1131 tracked `.md` (1027 after excluding evals fixtures + changelogs); tree clean; skill's literal default is a no-op on clean tree; detect.sh serial run over full corpus exceeds 5 min (needs chunking); `/session-flow:orchestrate` armed; user clarified intent mid-turn (orchestrated full-repo run + skill fallback design) -- [ ] Step 1.5: Auto-detect — user asked to be interviewed; forced Q&A (`me`-leaning) -- [ ] Step 2: Drive the frontier-rounds loop -- [ ] Step 3: Recognize the stop condition -- [ ] Step 4: Persist the contract -- [ ] Step 5: Hand off - -## Open-question register - -- Q1 | open | round 1 | Deliverable: report-only, or report + apply Tier 1 treatments in this session? | -- Q2 | open | round 1 | Corpus scope/exclusions for the repo-wide run? | -- Q3 | open | round 1 | Where does the audit report land (committed memory slice vs PR-only vs artifact)? | -- Q4 | open | round 1 | Orchestration shape: concurrency cap, model tier, verification pass? | -- Q5 | open | round 1 | Skill update scope: audit-noise fallback only, or propagate to sibling audit skills? | -- Q6 | open | round 1 | File the rate-limit-telemetry enhancement issue now, or defer? | - -## Decision tree (`me` mode) - -- [ ] Deliverable shape (Q1) -- [ ] Corpus scope (Q2) -- [ ] Report destination (Q3) -- [ ] Orchestration parameters (Q4) -- [ ] Skill fallback update scope (Q5) -- [ ] Rate-limit telemetry issue (Q6) - -## Session-shorthand glossary - -- "fallback" — the behavior `/docs-hygiene:audit-noise` should adopt when invoked with no target AND a clean tree (today: friendly no-op) -- "slices" — topic-docs working directories: `docs/topics//` (contract tier) and `.work//` (memory tier)