retro
Skill2 starsChanged 11 days agoMITView on GitHub
Review the sessions since the last retro in which skills ran, and log what each skill did wrong with evidence and a proposed fix.
194 other projects use a version of this.
Adopted by
Most recently taken up by
- icetomoyo/CodingSkillstheir own version, first seen Nov 16, 2026
- siarhei-belavus/agent-smitttheir own version, first seen Nov 16, 2026
- philoserf/claude-code-configtheir own version, first seen Nov 15, 2026
- jdonohoo/vern-bottheir own version, first seen Nov 15, 2026
- Dimon94/dveritytheir own version, first seen Nov 15, 2026
- Dimon94/skillstheir own version, first seen Nov 15, 2026
Used with
Projects with it also use plan-tune, ios-clean and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
ShreyRaviSkillplan-tune
Self-tuning question sensitivity + developer psychographic for gstack (v1: observational). Review which AskUserQuestion prompts fire across gstack skills, set per-question preferences (never-ask / always-ask / ask-only-for-one-way), inspect the dual-track profile (what you declared vs what your behavior suggests), and enable/disable question tuning. Conversational interface — no CLI syntax required.
Also in 9 projects that use this
Saqlain-811Skillios-clean
Remove the DebugBridge SPM package and all #if DEBUG wiring from an iOS app. (gstack)
Also in 5 projects that use this
Saqlain-811Skillios-fix
Autonomous iOS bug fixer. (gstack)
Also in 5 projects that use this
seek888Skillpair-agent
Pair a remote AI agent with your browser. One command generates a setup key and prints instructions the other agent can follow to connect. Works with OpenClaw, Hermes, Codex, Cursor, or any agent that can make HTTP requests. The remote agent gets its own tab with scoped access (read+write by default, admin on request). Use when asked to "pair agent", "connect agent", "share browser", "remote browser", "let another agent use my browser", or "give browser access". (gstack) Voice triggers (speech-to-text aliases): "pair agent", "connect agent", "share my browser", "remote browser access".
Also in 11 projects that use this
ShreyRaviSkilllanding-report
Read-only queue dashboard for workspace-aware ship. Shows which VERSION slots are currently claimed by open PRs, which sibling Conductor workspaces have WIP work likely to ship soon, and what slot /ship would pick next. No mutations — just a snapshot. Use when asked to "landing report", "what's in the queue", "show me open PRs", or "which version do I claim next". (gstack)
Also in 8 projects that use this
timurgaleevSkilldocument-generate
Generate missing docs for a feature, module or project as tutorials, how-tos, reference and explanation.
Also in 7 projects that use this
- +1 more from Saqlain-811
- +2 more from ShreyRavi
Other versions194 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 194
- mattpocock/skills★ 325ksame name
- garrytan/gstack★ 142ksame name
- formbricks/formbricks★ 13ksame name
- ArcReel/ArcReel★ 6.6ksame name
- vinvcn/mattpocock-skills-zh-CN★ 5.8ksame name
- withcoral/coral★ 4.9ksame name
- bergside/awesome-design-skills★ 2.9ksame name
- MiniMax-AI/minimax-code★ 2.2ksame name
- DenchHQ/DenchClaw★ 1.7ksame name
- DenchHQ/ironclaw★ 1.7ksame name
184 more not listed
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 repository22 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 2 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 169 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 22
Run digest.py --list --runs <run folder>, with --since <date> when the user gave one.
- Very specific: line 34
Read WATCH.md in the run folder, when it exists.
- Very specific: line 44
Read the findings in the run folder about the skills that ran.
- Distinctive: line 50
Look for these gaps in each timeline:
- Distinctive: line 102
Run file path · sessions read · findings: new, repeated, project · each watch item, with what the sessions showed or "not seen" · the …
“Common” is measured: kept by at least 6 in 10 of the 169 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
Retro
The skill set improves from evidence, not from memory, and each fix must keep it small. Run it about once a week. /retro reads each session since the last retro, in each project, compares what each skill did with what its SKILL.md says, and writes the gaps to one run file. It proposes fixes and edits no skill.
digest.py is in the folder of this skill. Run it with python3 -I.
1. Find the run folder
Resolve this SKILL.md with realpath. Walk up from that path, and look for a .git folder with ls. Run no git command on that path, because the worktree guard refuses it. When you find a repo other than the current project, that repo is the skills repo, and the run folder is runs/ at its root. When you do not, the run folder is .scratch/retro/ in the project. Tell the user to send the file to the owner of the skill set.
Completion: you know the run folder.
2. List the sessions
Run digest.py --list --runs <run folder>, with --since <date> when the user gave one. It prints one row for each session in which a skill of this set ran, after the lines that a run file already covers:
<session> · <project> · <skills> · lines <first>-<last> · <transcript path>With no row, tell the user that no session needs a retro, and stop.
Completion: you have the sessions, each with its line range.
3. Read the sessions
Read WATCH.md in the run folder, when it exists. Each of its items is one more gap to look for in each session, and each reader reports what the session showed for it.
For one or two sessions, read each one yourself: run digest.py <transcript> --from <first>. The output is a timeline: user messages, commands, skills, subagents, questions and tool errors, each with its transcript line number. Open the transcript at a line when you need the full text. Then find the gaps as step 5 says.
For three sessions or more, group the sessions by project, about 3 or 4 for each subagent, and start the subagents all at once. Give each one its rows, the path of this SKILL.md, the watch items, and this task: read the timeline, read the SKILL.md of each skill that ran, find the gaps as step 5 says, and return each gap in the finding format of step 6, with no status. Evidence cites <session>:<line>.
Completion: you have the gaps of each session.
4. Read earlier runs
Read the findings in the run folder about the skills that ran. Note each finding that is open and each that is rejected.
Completion: you have the open and rejected findings for each skill that ran.
5. Find and merge the gaps
Look for these gaps in each timeline:
- Skipped step: a step that did not run, or a completion criterion that the agent claimed but did not meet.
- Correction: the user corrected, repeated or stopped the agent.
- Wrong trigger: a skill fired when it should not, or did not fire when it should.
- Fact question: the agent asked the user for a fact that it could look up.
- Excuse: the agent gave a reason to skip work.
- Friction: a tool error, a retry, or a long detour.
- Outside skill: a skill or agent that is not in this set ran. Record what it did that the set did not do.
For each gap, decide the fix: cut a step or a no-op, sharpen the wording, change a description, add a step, or, for an outside skill, add its idea to a skill of this set. A fix must remove at least as many lines as it adds from the skill, and a skill stays at about 80 lines. When a gap has no fix that fits, propose what to cut with it, or drop the finding. A gap that is caused by the project and not by a skill gets the project tag and no fix.
Merge the gaps of all sessions. The same gap in more than one session is one finding, with evidence from each session. When a gap matches an open finding from step 4, link that finding and count the runs. When it matches a rejected finding, leave it out.
Completion: each finding has a skill and step, evidence with <session>:<line>, and a fix or the project tag.
6. Confirm and write
Show the findings to the user, ordered by how many sessions and runs saw them, then by cost to the user. Give each finding its own line in the chat, with its ID, skill and title, so the user can drop or change each one without opening the file. A count in place of the lines is not a list. Then write <run folder>/<date>-<projects>-<skills>.md in this format:
# <date> · <projects> · <skills that ran>
<One or two sentences: what the sessions did.>
## Sessions
- <transcript path> · lines <first>-<last>
## Findings
### F1 · <skill> · step <n> · <short title>
Status: open
Seen in: <the sessions of this retro, and each earlier run file and finding number, with the count, or "first time">
What happened: <one or two sentences>
Evidence: <session>:<line>: "<quote>"
Fix: <fix type>: <the change to make>
## Worked
- <a skill step that did what it should, worth keeping as it is>The Sessions list is the coverage record: the next /retro starts each transcript after its last line. Write each row from step 2, with the path in full and the range unchanged.
A finding's ID is its file name and number, for example 2026-10-15-habits-spec#F1. The commit that applies a fix names that ID in its message and sets Status: applied in the same change. git log --grep <ID> then finds the fix. When we decide against a fix, set Status: rejected: <reason>.
Completion: the run file exists, its Sessions list has each row from step 2, and the user approved its findings.
Report
Run file path · sessions read · findings: new, repeated, project · each watch item, with what the sessions showed or "not seen" · the finding to fix first.
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.

