Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add UnboundCompute/security-agent-skills --skill hunting-bug-variantsgit clone --depth 1 https://github.com/UnboundCompute/security-agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/unboundcompute/security-agent-skills/hunting-bug-variants)<a href="https://agentmods.dev/skills/unboundcompute/security-agent-skills/hunting-bug-variants"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/hunting-bug-variants/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/unboundcompute/security-agent-skills/hunting-bug-variants"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/hunting-bug-variants.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00121 | $0.01596 |
| Opus 5.5 | $0.00048 | $0.00638 |
| Sonnet 5.5 | $0.00024 | $0.00319 |
| Haiku 4.5 | $0.00012 | $0.00160 |
Grade A, and why
hunting-bug-variants scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 133 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hunting bug variants: one seed, every sibling
One bug is almost never alone. The same mistake gets copy-pasted, re-implemented by a second author, or left uncovered by a fix that only patched the reported call site. Variant hunting takes a single confirmed defect and turns it into a signature you can sweep the whole codebase with, so you find the three other copies before an attacker does.
When to use
- You just confirmed a bug and want its siblings, not a victory lap.
- A CVE, advisory, or writeup describes a bug in code you can read.
- A fix landed and you want to know whether it covered every instance.
- Someone says "we already patched that" and you want to verify it is really gone.
Scope check
Authorized source only (your own, OSS, CTF, in-scope engagement). If you can't name the authorization, stop.
The loop
-
Anchor the seed. State the confirmed defect precisely: the exact sink, the exact missing or wrong step (no bounds check, no authz check, unnormalized path, unescaped context), and the untrusted input that drives it. A vague seed ("there was an injection somewhere") produces a useless sweep.
-
Extract the shape, not the string. Abstract the seed into a structural signature that survives renaming and reformatting. Usually one of: this sink called without that guard; this unsafe idiom (raw concat into a query,
base / user_namewithout resolve-and-contain); this type of value reaching this argument. Grepping the literal code finds only exact copies; the shape finds the re-implementations. -
Sweep for same-shape sites. Enumerate every place matching the signature: other callers of the same sink, other functions with the same step missing, copy-paste clones, sibling handlers in the same module, the same pattern in a different subsystem. Cast wide here; ranking is triage, not a filter.
-
Check the fix itself for gaps. If the seed came with a patch, treat the fix as suspect: does it cover every reachable path to the sink, or only the reported one? Was the check added to the caller but not a second caller? Is there a pre-existing path that bypasses the new check? Incomplete fixes are the highest-yield variants because everyone believes the bug is dead.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 133 lines · 121 tokens per session scan A 05f2d04d1121
hunting-bug-variants is a skill published in the GitHub repository UnboundCompute/security-agent-skills (5 stars, last pushed 1mo ago), licensed MIT. It adds 121 tokens to every session and 1,596 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-05.
Other skills, from other repositories
ffuf-claude-skill
Skill "ffuf-claude-skill" from Agent-Threat-Rule/agent-threat-rules, covering ffuf claude skill, when to use this skill and instructions.
codeinspectus-fix-one
Investigate and remediate exactly one user-selected CodeInspectus finding with evidence-gated reproduction, a separately approved minimal patch, focused regression testing, and an exact-prior-scan rescan. Use when a user asks an agent to examine, reproduce, fix, or verify one CodeInspectus finding without batching…
gauntlex:doctor
Full GAUNTLEX environment health check — model reachable, ChromaDB writable, AVF fixtures pass, no unexpected outbound network (air-gap verification). Use before first run or when debugging failures.
report-error
Verify a suspected plugin defect locally, prepare a minimal anonymised GitHub issue, and publish only after the user reviews and explicitly approves its exact content. Also supports a local-only diagnostic bundle.
prodcheck-review
Review this codebase against the prodcheck pre-production checklists — security, performance, scale, integrations, post-launch readiness and people. Use when asked to check whether a project is ready to ship, to audit an area before launch, or to work through a specific checklist. Produces evidence with file:line…
diagnose-bundle
Maintainer/dev skill that triages an anonymised diagnostic bundle (appsec-diag-.tgz, produced by scripts/runtime/diagnosticbundle.py) a user sent after a pipeline failure. Runs the deterministic inspect, then cross-references the plugin source (scripts/, compact runtimes, agents/, AGENTS.md) and known-bug history to…