agentleFS
mtarcuremtarcure/claude-vibe-squad/.agents/skills/dedup-prior-art-check/SKILL.md

dedup-prior-art-check

Skill220 starsChanged 20 days agoMITView on GitHub

Use before spending real effort on a bug-bounty lead, and again immediately before submitting a finding, to prove the vulnerability and its specific composite chain are not already known — public advisories, prior audits, Solodit, and our own past submissions — so the squad never burns a report on a duplicate. Fires at systematic-attacking Phase 1 and its pre-submit refresh.

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:

    Before spending effort on a lead and again immediately before submitting a finding, prove the vulnerability (and the specific composite …

    line 9
  2. Distinctive:

    Every lead carries a dated novelty verdict with the prior-art sources queried named.

    line 41
  3. Distinctive:

    Every lead carries a dated novelty verdict before deep effort.

    line 42

≈ “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

Dedup / Prior-Art / Novelty Check

Before spending effort on a lead and again immediately before submitting a finding, prove the vulnerability (and the specific composite chain) is not already known — so the squad never burns a report on a duplicate. This operationalizes systematic-attacking Phase 1 and its pre-submit refresh into a concrete, evidence-producing gate.

Source: corpus B §11 (Cyfrin Solodit — 49k+ aggregated real-world smart-contract findings) plus public advisory databases (GHSA, CVE, OSV) and program disclosure history; corpus A elite-vs-average (narrow-and-deep continuous prior-art discipline). Impact class enabler: protects report acceptance rate; a duplicate/likely-duplicate verdict routes a lead for differentiation. This gate does not itself kill or demote - campaign disposition, and the single Phase-5 candidate-kill point, is owned by shared/modes/bounty.md. Governing method: the dedup gate of systematic-attacking (Phase 1 + pre-submit).

Method

  1. Build the search key from the lead: root-cause class, affected contract/endpoint/component, protocol + version, and the specific chain shape (not just the vuln class).
  2. Query the prior-art surfaces: program disclosure history (HackerOne/Immunefi), GHSA/CVE/OSV, and — for web3 — the Solodit corpus. Use a separately provisioned prior-art service when one is configured for this lane; otherwise search Solodit and the GitHub advisory database by hand. Log a prior-art service attempt that fails as needs_tool, even when the manual search then succeeds. Search our own chrono-vault for prior squad work on the target.
  3. Classify: novel / variant-of-known (cite the closest prior finding + what differs) / likely-duplicate / duplicate. Composite chains get a separate verdict — a chain of known primitives can still be a novel composition, and a novel-looking chain can be a known composite.
  4. Return the novelty verdict with evidence; do not kill or demote - that is the campaign's call (shared/modes/bounty.md, Phase 5). An exact collision leaves the active queue but is preserved as an inner-link primitive for the chainer ('differentiate, don't veto'). For variant-of-known, document the delta that makes it independently submittable.
  5. Refresh the check immediately before submission (new disclosures land daily); attach the dated query evidence to the report.

Acceptance

  • Every lead carries a dated novelty verdict with the prior-art sources queried named.
  • Every lead carries a dated novelty verdict before deep effort; disposition (kill/demote) is the campaign's Phase-5 decision, not this gate's, and exact collisions are preserved as chain primitives.
  • The pre-submit refresh is recorded with its date; a failed prior-art service attempt is logged as needs_tool — even when the manual search succeeded — never silently skipped.

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.