agentleFS
mtarcuremtarcure/claude-vibe-squad/.agents/skills/review-severity-ladder/SKILL.md

review-severity-ladder

Skill220 starsChanged 20 days agoMITView on GitHub

Use when you need severity to mean the same thing across reviewers, specialists, and models — place each finding on one shared critical/high/medium/low/note rung by consequence rather than fix effort, keep severity separate from confidence, and declare the deliverable's reporting floor.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Place each finding on the ladder by consequence, not by effort to fix and not by how interesting it is:

    line 12
  2. Distinctive:

    When the ladder produces repeated critical findings on the same mechanism, the correct move is to remove or bound the mechanism, not to …

    line 23
  3. Distinctive:

    Every finding carries exactly one rung and a one-line consequence statement.

    line 26

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Review Severity Ladder

Rank findings on one shared ladder so severity means the same thing across reviewers, specialists, and models.

Steps

  1. Place each finding on the ladder by consequence, not by effort to fix and not by how interesting it is:
    • critical — data loss, credential exposure, unauthorized state change, or silent corruption reachable on a normal path.
    • high — wrong results, broken invariant, or availability loss on a reachable path; correct behavior depends on luck.
    • medium — wrong behavior on an edge, boundary, or error path; degraded but recoverable.
    • low — quality, clarity, or duplication that will cause a future defect but causes none now.
    • note — observation with no defect claim.
  2. Justify each placement with the consequence and the path that reaches it. "Could be a problem" is not a placement.
  3. Apply the reachability test before assigning critical or high: name the entry point, the actor, and the input that gets there. Unreachable code caps at medium.
  4. Apply the intrinsic-impact test: a finding that only reveals information already disclosed, or that requires privileges the attacker would not have, does not clear high.
  5. Separate severity from confidence. A high-severity finding you are unsure of is high / unconfirmed — never demote severity to express doubt, and never inflate confidence to justify severity.
  6. Set a severity floor for the deliverable and report everything at or above it; list the rest as notes rather than discarding them.
  7. When the ladder produces repeated critical findings on the same mechanism, the correct move is to remove or bound the mechanism, not to keep escalating the threat model.

Acceptance

  • Every finding carries exactly one rung and a one-line consequence statement.
  • Critical and high findings name entry point, actor, and reaching input.
  • Severity and confidence are recorded as separate fields.
  • The deliverable states its severity floor.
  • Placements are defensible without reference to how hard the fix is.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.