Skip to content
Agentmart
Submit a tool

maintain-formal-verification-skill

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…

Skill by kevherro

IndexedMIT
Write a review

Chalk Labs

Need help deploying maintain-formal-verification-skill?

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.

Book a call

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-maintain-formal-verification-skill

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.

We use your email only for this newsletter. Privacy

About

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 each failure as drift or bug. Use for /maintain-formal-verification-skill, "does the model still match", or after merges into covered code that never reran the check.

No 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 1, 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/maintain-formal-verification-skill

What's inside

Skill name
maintain-formal-verification-skill
Description
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 each failure as drift or bug. Use for /maintain-formal-verification-skill, "does the model still match", or after merges into covered code that never reran the check.
License
MIT
Instructions
84 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Maintain a formal verification skill

A model that has fallen behind the code can stay green while
describing functions that are gone. That green result reads like a
proof, so it is the case to fear. This skill handles changes that
bypassed the project skill: merges, edits by agents that never loaded
it, and slow drift over many commits.

Location and the generated-skill contract live in the discovered
**create-formal-verification-skill** skill. Do not restate them.

## Outcome per check

- **in sync.** Nothing since the last sync alters what the check
  models, and it passes.
- **updated.** The model now matches the new code, and it passes.
- **bug.** The model matches current code and the code violates the
  property. That is a code finding, not drift.
- **blocked.** It could not be re-synced or could not run. Give the
  concrete reason.

One check may end both updated and bug.

## Scope

Touch only verification files: models, harnesses, properties, and the
project skill. Product code only on the user's request. Never relax a
property or model to turn a failure green.

## Pass

1. **Find the checks.** Look for `formal-*` skills, then any project
   skill that drives a model checker, prover, or verifier against
   named code. Per check, note the property, covered code, bounds,
   command, and sync commit. No sync commit recorded: take the last
   commit touching the model files, and say so.
   Nothing found: stop and point at
   `/create-formal-verification-skill`.
2. **Run as-is.** Record the result before editing. Green here only
   says the old model still holds. Do not classify a red result yet.
3. **Read every covered change.** Diff covered files from the sync
   commit to `HEAD` and read the messages. Follow renames, splits,
   and moves. Search for new code that writes the guarded state.
   Classify each change:
   - **No effect.** Judged against the model, not the commit
     message.
   - **Shape changed.** An atomic step split, a branch added, a check
     moved, a new route to the guarded state, a widened value range.
     The model must change.
   - **Promise changed on purpose.** The change itself says so (doc
     comment, tests, changelog, message). The property changes, citing
     that evidence.
   - **Promise maybe broken.** The code moved with no such evidence.
     A later failure is a bug candidate.
4. **Update from the code.** Write the model edit from the code
   before rerunning, not from what turns it green. Keep atomic steps
   atomic. Edit a property only for an intended promise change, and
   record old wording, new wording, and evidence. Keep bounds unless
   the change invalidates them; report any reduction.
5. **Rerun and judge.**
   - **Green.** Sabotage the new code in the way the property is
     meant to catch and confirm the check goes red. Undo the
     sabotage.
   - **Red.** Walk the counterexample through real code one step at
     a time. A step the code cannot take means the model is wrong;
     fix it and say what was wrong. If every step is reachable, it is
     a bug: name the introducing commit, the steps in the code's
     words, and the harm. A cheap repro against real code is report
     evidence, not a product commit.
6. **Update the skill.** Coverage, commands, bounds, assumptions,
   and the sync commit, set to the commit you verified against. A
   blocked check keeps its old sync commit. Then walk the skill once
   as the next agent would.

**Reply:** per check, the outcome and the commits read; each model
edit and why (shape change, or intended promise change plus
evidence); the rerun plus the sabotage that proved it can still fail;
each bug as steps, harm, and introducing commit; remaining coverage
gaps. Separate drift from bugs explicitly. If a check was green
before the update and red after, say the earlier green came from a
stale model. Verification edits go in their own commit, apart from
any product fix, and only if the user asked for commits.

From kevherro/mestack, MIT.

More from kevherro/mestack

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
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from kevherro

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 maintain-formal-verification-skill 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.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying maintain-formal-verification-skill?

CallGet help
Cite this page

maintain-formal-verification-skill. Agent Mart. https://agentmart.dev/skills/kevherro-maintain-formal-verification-skill (last updated 2026-10-24)