Repository navigation
docs(skills): point lifecycle SKILL.md at the counts check.sh prints - #338
Conversation
Replace the hand-written run, guard, witness, state and CI share counts in the verify-session-lifecycle SKILL.md with pointers to the lines that check.sh prints. check.sh now prints the run count and the count of runs that expect pass before TLC starts, so CHECK_JOBS can be set to that count.
|
@codex review |
|
@codex security review |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Codex Review: Didn't find any major issues. Already looking forward to the next diff. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
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". |
🛡️ Codex Security ReviewSecurity review completed. No security issues were found in this pull request. Reviewed commit: Only the user who started this review can view the report in Codex. ℹ️ About Codex security reviews in GitHubThis is an experimental Codex feature. Security reviews are triggered when:
Once complete, Codex will leave suggestions, or a comment if no findings are found. |
Drop the hand-written share, run, vCPU and minute counts from the comment above the lifecycle-model job. The matrix values stay as config.
|
@codex review |
|
@codex security review |
|
Codex Review: Didn't find any major issues. Another round soon, please! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
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". |
🛡️ Codex Security ReviewSecurity review completed. No security issues were found in this pull request. Reviewed commit: Only the user who started this review can view the report in Codex. ℹ️ About Codex security reviews in GitHubThis is an experimental Codex feature. Security reviews are triggered when:
Once complete, Codex will leave suggestions, or a comment if no findings are found. |
Problem
The verify-session-lifecycle
SKILL.mdheld hand-written counts of facts thatcheck.shcomputes or that change when a run is added: the run count ("94 times",CHECK_JOBS=94), the runs over 20 million states ("Fifteen runs"), the guard count ("forty guards"), the no-token witness count ("Eight of those states"), the distinct states ofcurrentandcurrent-no-token, and the CI share count and minutes ("nine jobs",CHECK_SHARD=<i>/9,CHECK_JOBS=12, "11 to 33 min"). Every model PR had to edit them. Parallel PRs conflicted on those lines (#330 and #331 both edited theCHECK_JOBSline), and the numbers went stale (6,954,727 states written whilecheck.shprinted 8,609,456).Change
SKILL.mdnames no count. Each count is now a pointer to the line thatcheck.shprints:Runs:,Facts:,Distinct states of current:,Distinct states of current-no-token:andShare <i>/<n> of the runs:. The CI paragraph points to thelifecycle-modelmatrix in.github/workflows/ci.ymlfor the share count andCHECK_JOBS.check.shprints one new line before TLC starts:Runs: <n>, <p> of them expect pass.On main it printsRuns: 94, 17 of them expect pass.WithCHECK_SHARDit counts the share.CHECK_JOBS: the skill now says "CHECK_JOBSset to that run count". I chose this over a newCHECK_JOBS=allvalue because it is the smaller change (oneecho, no new input to validate).coverage.json: only theskill:scripts/check.shfingerprint changed, becausecheck.shis covered..github/workflows/ci.yml: the comment abovelifecycle-modelno longer names counts ("the 9 shares", "of the 94 runs", "4 vCPUs", "11 to 33 min"). It points to theRuns:,Share <i>/<n> of the runs:andCPUs:lines ofcheck.sh. The matrix values (shard,CHECK_JOBS,CHECK_SHARD) stay, because they are config.actionlintandzizmor --offline .are clean.The model, the guards and the expected results are unchanged. The stable facts stay (TLC 1.7.4 and its sha256, rule names, code constants, measured timings).
Validation
check.sh --drift-only: "Covered code matches the sync record." (exit 0).CHECK_SHARD=9/9 CHECK_JOBS=12 check.sh --update-coveragewith real TLC: all 9 runs as expected, exit 0, and the new line printedRuns: 9, 1 of them expect pass.check.shwith a fakejavaand no shard:Runs: 94, 17 of them expect pass.oxfmtandnpm run format:check: clean.lifecycle-modeljob runs every share, becausecoverage.jsonchanged.Generated with
xum• Model:anthropic:claude-opus-5-5• Thinking:high