Conversation
…ned generations Finding 5 of #1383's round-1 audit, left unfixed when that PR merged. Recorded plainly: I reported four findings to the operator and fixed four, and this was a fifth (🟡) I did not carry over. #1383 shipped timestamped plan generations so a bad save no longer destroys the previous one, and it printed the recovery command on stderr. But the operator- facing doc for this workflow — the SKILL that auto-loads on trigger — still described `save` as writing two files, with no mention of generations or `restore --plan <generation>`. Measured before this change: zero hits for either term in claude/skills/initiative-scan/SKILL.md. That matters for exactly the case the feature exists for. After a bad save an agent reads the skill, not scripts/README.md (which #1383 did update and which nothing auto-loads), and would learn nothing about the recovery path. On 2026-09-07 a continuum autosave overwrote a 47-entry plan with the 10-entry post-crash workspace and 20 windows had to be identified by hand. Every claim in the added text was verified against the SHIPPED code rather than copied from the PR body: KEEP_GENERATIONS = 192, GENERATIONS_DIRNAME = "restore-plans" (192 x 15min = 48h) `--plan` bypasses the staleness gate — tmux-session-restore.py:1353, `if staleness_hours is not None and plan_path is None:` stamps are UTC — `time.gmtime` in generation_stamp live sample on this host: restore-plan_20260910T012436.json Skill-listing and tier gates re-run: 61 passed. This adds body text only, which costs nothing until the skill is invoked; the always-on `description` is unchanged. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013jdbmhCKa6edhTmiADsziR Claude-Session-Id: 097b404c-db17-4472-bd37-dc90cf8fa675
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Finding 5 of #1383's round-1 audit, unfixed when that PR merged — I reported four findings and fixed four; this was a fifth (🟡) I did not carry over.
#1383 shipped plan generations and prints a recovery command on stderr, but the skill that auto-loads for this workflow still described
saveas writing two files. Measured before the change: 0 hits forgenerationorrestore --planin that file. After a bad save an agent reads the skill, notscripts/README.md(which #1383 did update, and which nothing auto-loads).Every claim in the added text was verified against the shipped code, not copied from the PR body — including
--planbypassing the staleness gate (tmux-session-restore.py:1353) and stamps being UTC. Skill-listing + tier gates: 61 passed. Body text only; the always-ondescriptionis unchanged.