verify
Skill1 starsChanged 32 days agoMITView on GitHub
Project recipe for /verify in retroagent. Drive this checkout's kb CLI and plugin hooks end to end in a sandbox data repo with drive.py (kb commands, or headless Claude Code with the plugin), then judge by the evidence it prints. Use when you verify a change to retroagent.
- Installs packages
188 other projects use a version of this.
Adopted by
Most recently taken up by
- Doist/todoist-aitheir own version, first seen Nov 9, 2026
- udecode/better-convextheir own version, first seen Nov 9, 2026
- pytorch/heliontheir own version, first seen Nov 8, 2026
- amir20/dozzletheir own version, first seen Nov 8, 2026
- Doist/todoist-clitheir own version, first seen Nov 8, 2026
- codewithmukesh/dotnet-claude-kittheir own version, first seen Nov 8, 2026
Used with
Projects with it also use instinct-export, instinct-import 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.
TheKREATORSSkillinstinct-export
Export instincts from project/global scope to a file
Also in 3 projects that use this
TheKREATORSSkillinstinct-import
Import instincts from file or URL into project/global scope
Also in 3 projects that use this
AviadPSkillrisk-assess
Assess the risk and blast radius of a proposed code change before implementing it. Evaluates CI impact, framework-wide vs localized scope, rollback path, and identifies blindspots. Use when user says "what's the risk", "is this safe", "assess the impact", "risk assess", "blast radius", or before applying framework-wide or CI-affecting changes. Do NOT use for code quality review (use code-quality) or plan validation (use design-validator).
Also in 3 projects that use this
diego-tobalinaSkillbuild-fix
Fix build and compilation errors systematically. Run when build fails.
Also in 6 projects that use this
AaravChadhaSkillsecure
Security review with a confidence gate - a finding exists only with a concrete exploit scenario and a high/medium/low confidence rating; everything else goes to a worth-hardening list, never inflated. Sweeps five surfaces: auth gates, secrets hygiene, injection and unsafe sinks, unsafe deserialization/crypto/transport, and LLM tool-use trust boundaries. Reports only, never fixes. Use when the user asks for a security review or to check vulnerabilities, secrets, or auth.
Also in 3 projects that use this
ArvorCoSkillspecify
Create feature specification from natural language description. Use when starting a new feature or creating a spec. Triggers on: create spec, specify feature, new feature spec.
Also in 10 projects that use this
- +1 more from TheKREATORS
Other versions188 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 188
- react/react★ 252ksame name
- facebook/react★ 252ksame name
- garrytan/gstack★ 141ksame name
- oven-sh/bun★ 97ksame name
- anthropics/financial-services★ 41ksame name
- anthropics/financial-services-plugins★ 41ksame name
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- anthropics/claude-quickstarts★ 19ksame name
- NVIDIA-NeMo/Speech★ 19ksame name
178 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.
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 167 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 57
No change in data.log after a hook run with auto_sync: false.
“Common” is measured: kept by at least 6 in 10 of the 167 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
Verify retroagent
Two surfaces, both against one sandbox data repo:
- The CLI. This checkout's
bin/kb. - The hooks. Real headless Claude Code with this checkout as the plugin: SessionStart (
kb-hook, the brief), PreToolUse (kb-guard), PostToolUseFailure (kb-hint), and the decide mod'ssession.start.
Report SKIP when the diff touches only tests, docs or scripts/pages-routine.md: there is nothing to drive. Codex
runs the same hook scripts; it is not driven here.
drive.py (next to this file, stdlib Python) makes the sandbox, runs the drive, prints the evidence and cleans up.
--help says what the sandbox holds and how runs stay off the real data repo. Never run bin/kb by hand for
/verify: without the sandbox's KB_CONFIG, it reaches the real data repo.
Pick the drive
| Changed | Drive |
A kb command (src/kb/) | kb with that command |
| The sync, gitops, redaction | new --no-sync, then kb <dir> -- sync --now --no-summaries; judge data |
bin/kb-hook, the brief | hook with any prompt; read the SessionStart answer in hookEvents |
bin/kb-guard | hook with a prompt that runs the command to deny or allow |
The hand-over of a cached copy to the code clone (bin/kb, bin/kb-guard) | new --config '{"code": "<clone>"}', then hook --cached; pluginLog shows the hook path under <dir>/plugins/cache/ |
bin/kb-hint, src/kb/hint.py | hook with a Bash command that fails with an error the seed page knows |
| A skill | hook with the slash command as the prompt (/retroagent:kb-search flaky test) |
hooks/decide.tsx | claude plugin validate . and claude plugin test .; hook shows the mod load and its kb decide --json run in pluginLog |
Commands
python3 .claude/skills/verify/drive.py new # prints dir
python3 .claude/skills/verify/drive.py kb <dir> -- find flaky
python3 .claude/skills/verify/drive.py hook <dir> 'Run exactly this Bash command: ...' '<second prompt>'
python3 .claude/skills/verify/drive.py clean <dir> # after you read the evidenceMake one sandbox per /verify and pass its folder to each run. Judge by hookEvents, pluginLog, hintLog and
data, never by the reply alone. One Haiku turn costs about $0.01.
Probes
Checked on 2026-10-13, all from one sandbox:
kb <dir> -- find flaky→ thedemo-apppage and session55555555.- SessionStart → the page pointer,
[s-000001], and "1 retro suggestion waits". Run exactly this Bash command: kb decide accept s-000002 --yes→kb-guarddenies it.hook --cachedwith"code"set to a folder whosebin/kb-guardonly prints a deny with its own reason →echo hellois denied with that reason, so the cached guard ran the clone's. With"code"set to this checkout, thekb decide acceptprobe is denied andechogoes through.Run exactly this Bash command: sh -c 'echo "ModuleNotFoundError: No module named yaml" >&2; exit 1'→kb-hintadds thepip install -e .[dev]fix, andhintLoghas one line.- No change in
data.logafter a hook run withauto_sync: false.
Gotchas
- Your own installed
kb-guardalso checks your Bash calls. A copy older than the tokenizing guard (PR #49) denies a drive command that holdskb decide accept|reject, even in quotes. Runkb updatefirst. - Managed hooks run too (
--setting-sources project,localdrops only user settings), sohookEventscan hold their answers.pluginLognames the retroagent hooks (bin/kb-hook,bin/kb-guard,bin/kb-hint). - A
kbrun gets the sandboxHOME; a hook session keeps the realHOMEfor the login. Sokb-hintin a hook can use the local embedding model (method: semantic), and akb hintrun is keyword only. - The SessionStart sync starts only with
new --config '{"auto_sync": true, "debounce_minutes": 0}': the first sync ofnewwriteslast-ok. The sync runs in the background after the hook. Readkb <dir> -- statuslater. - The model can change a command or skip a step. Phrase prompts as "Run exactly this ...". If a step did not happen, run again before you call it a failure.
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.

