retrospect
Skill1 starsChanged 5 days agoMITView on GitHub
Use after delivery, release, incidents or long unattended runs to turn evidence into process changes and regression gates. Not for context cleanup (session), skill audits (skill-craft) or active defects (diagnose, implement).
3 other projects use a version of this.
Adopted by
Most recently taken up by
- grove/promise-to-prooftheir own version, first seen Oct 12, 2026
- ngc-shj/claude-code-configtheir own version, first seen Aug 31, 2026
- npow/claude-skillstheir own version, first seen May 8, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- grove/promise-to-proof★ 4same name
- npow/claude-skills★ 4same name
- ngc-shj/claude-code-config★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository21 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- code-reviewSkill
- deliverSkill
- diagnoseSkill
- engineering-docsSkill
- engineering-writingSkill
- evaluateSkill
- implementSkill
- legal-writingSkill
- natural-writingSkill
- orchestrateSkill
- refactorSkill
- releaseSkill
- scopeSkill
- sessionSkill
- setup-repoSkill
- skill-craftSkill
- test-designSkill
- trackerSkill
- ui-ux-designSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Read what exists, and mark what is absent unknown rather than inferring success from a clean final answer:
- Very specific: line 25
Test-suite health: tests added against removed, the mock ratio and assertion changes.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Retrospect
Change the process only on evidence, and only by the smallest change that
would have prevented the observed cost. A clean routine cycle returns
no change. One dated log per run is clutter, not a retrospective.
Evidence
Read what exists, and mark what is absent unknown rather than inferring
success from a clean final answer:
provenant telemetry stats --by preset(orpr,issue,role) for attempts, cost and retries;provenant telemetry tracefor one run's path;provenant telemetry experiments show NAMEfor arm comparisons.provenant quota usageandprovenant quota historyfor spend against allowance.- Review and repair history, escaped defects, the PR threads and any human correction of an unattended run. A correction names a class of defect: sweep for the class.
- Test-suite health: tests added against removed, the mock ratio and
assertion changes. Use the target repository's test-diff tooling where its
AGENTS.md,CONTRIBUTINGor runbooks name it; otherwise compare the test diff by hand. A suite that only grows is a finding.
For a wide review, fan out readers with provenant run --use explore --route junior --over costs,quality,recovery --file Q.md, one slice of
runs per branch.
Method
- Compare the outcome with the declared criteria and with a comparable run.
- Cluster recurring failures by cause: spec, test, skill, instructions, routing, tool, context or authority.
- Propose per cluster the smallest change, with evidence, owner, risk, destination and a success measure.
- Turn a representative failure into a deterministic test or a fixture before claiming improvement.
- Watch the next comparable cycle for recurrence and new costs.
improvedneeds a passing regression gate and enough comparable later runs; underpowered or confounded evidence isinconclusive.
Once a quarter, refresh the research behind the skills against 2026 primary sources and mark older ones as possibly stale.
Output
Proposal-first and read-only. Apply changes only under explicit authority or
inside an implement run. Finish with a table: finding, evidence, cause,
change, regression gate, destination, status (promote, experiment,
defer, reject). Promote durable conclusions to their owner (spec, ADR,
runbook, instructions, test, skill) and merge with existing text; never keep a
parallel diary.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.