correct
Turn the mistakes agents repeat in this repo into ones they cannot make, preferring the strongest enforcement that fits.
Skill by kevherro
Chalk Labs
Need help deploying correct?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/kevherro-correct
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Nov 2, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/correct
What's inside
- Skill name
correct- Description
- Turn the mistakes agents repeat in this repo into ones they cannot make, preferring the strongest enforcement that fits. Use for /correct, "stop making this mistake", or after the user corrects the same thing twice.
- License
- MIT
- Instructions
- 47 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Correct
The user keeps catching the same agent mistakes here. Change the repo
so the next agent cannot make them.
Assume the next contributor is an agent that read one file, copied
the nearest example, and took the shortest path that compiles. A
change that looks right from one file should be right for the repo.
## Steps
1. Collect evidence. Recent commits and reverts, review comments,
agent instruction files, workaround comments, and this session's
corrections. Group them into mistake classes. A class needs two
occurrences.
2. For each class, pick the highest level that works
(**encode-the-rule**):
1. **Structure.** One owner per piece of state, one supported way
per task. Make internals private so reaching for them breaks
the build. Generate lists that people sync by hand. Delete old
paths an agent would copy
(**migrate-callers-then-delete-legacy-apis**).
2. **Types.** Make the bad state unwritable
(**type-system-discipline**). If wrong code still builds, add a
lint or CI rule whose message tells the agent what to use. If
the pattern is already widespread, fail only on new
occurrences.
3. **Tests.** Fix or delete tests that miss the behavior they
claim to cover (**test-behavior**).
4. **Docs or agent rules.** Last, for judgment calls only. Nothing
fails when an agent skips them.
3. Land the most frequent classes now, one commit each if the user
asked for commits. Show each new guard rejecting an actual
historical instance (check out the old code, or replay the bad
diff). Run the same command locally that CI runs.
4. Exceptions live on the offending line with a reason and an
expiry. Granting one is the user's call.
5. Keep a rule table in the repo's agent instruction file: rule,
and what enforces it. A rule that is already listed with nothing
enforcing it is a repeat. Move it up a level in the same change.
Retire a rule when its mistake is no longer possible.
**Reply:** per class, the occurrences that prove it, the level
chosen, why nothing stronger fit, and the command that shows the
check failing on the old mistake.From kevherro/mestack, MIT.
More from kevherro/mestack
- 0.567score
mestack
kevherro
Kevin's harness-agnostic engineering dispatcher.
SkillAFreshness: AClaude Code +9 - 0.567score
create-verification-skill
kevherro
Generate a project-local verify-<app> skill that drives this repo's real app the way a user does.
SkillAFreshness: AClaude Code +9 - 0.565score
create-formal-verification-skill
kevherro
Write a repo-local formal-<area> skill that machine-checks one high-stakes promise tests rarely exercise, using the verifier that fits the question.
SkillAFreshness: AClaude Code +9 - 0.565score
maintain-formal-verification-skill
kevherro
Re-sync a repo's formal-* checks with code that moved since their last confirmation: read those commits, update the model from the code, rerun, and classify…
SkillAFreshness: AClaude Code +9 - 0.563score
interrogate
kevherro
Adversarial review of a change by one child per configured reviewer.
SkillAFreshness: AClaude Code +9 - 0.561score
benchmark-checklist
kevherro
Vet a performance measurement before you report or act on it: limiter, tuning, physical limits, errors, repeatability, relevance, and whether the work ran.
SkillAFreshness: AClaude Code +9 - 0.561score
ketchup
kevherro
Bring the user current on this chat since they last wrote, plainly, then lay out each pending call of theirs with what they need to make it without clicking…
SkillAFreshness: AClaude Code +9 - 0.561score
principle-attack-the-premise
kevherro
Two fixes on one belief failed the same check; test the belief before a third.
SkillAFreshness: AClaude Code +9 - 0.561score
help
kevherro
Print a table of the current mestack skills and a short description of each.
SkillAFreshness: AClaude Code +9 - 0.556score
in-your-own-words
kevherro
Play back the user's goal and the problem behind it before any work.
SkillAFreshness: AClaude Code +9 - 0.521score
figure-it-out
kevherro
Design and run an auditable playbook when no narrower mestack playbook fits.
SkillBFreshness: BClaude Code +9 - 0.517score
nudge
kevherro
Recurring check-in in this session: ask whether to write a paste-ready team update.
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 2
- Open issues
- 0
- Last commit (this folder)
- Oct 23, 2026 (11 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/kevherro/mestack
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from kevherro
- 0.567score
mestack
kevherro
Kevin's harness-agnostic engineering dispatcher.
SkillAFreshness: AClaude Code +9 - 0.567score
create-verification-skill
kevherro
Generate a project-local verify-<app> skill that drives this repo's real app the way a user does.
SkillAFreshness: AClaude Code +9 - 0.565score
maintain-formal-verification-skill
kevherro
Re-sync a repo's formal-* checks with code that moved since their last confirmation: read those commits, update the model from the code, rerun, and classify…
SkillAFreshness: AClaude Code +9 - 0.565score
create-formal-verification-skill
kevherro
Write a repo-local formal-<area> skill that machine-checks one high-stakes promise tests rarely exercise, using the verifier that fits the question.
SkillAFreshness: AClaude Code +9 - 0.563score
interrogate
kevherro
Adversarial review of a change by one child per configured reviewer.
SkillAFreshness: AClaude Code +9 - 0.561score
benchmark-checklist
kevherro
Vet a performance measurement before you report or act on it: limiter, tuning, physical limits, errors, repeatability, relevance, and whether the work ran.
SkillAFreshness: AClaude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get correct running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
correct skill. Agent Mart. https://agentmart.dev/skills/kevherro-correct (last updated 2026-10-24)