agentleFS
zyphr-labszyphr-labs/bounty-harness/.claude/skills/ai-llm/SKILL.md

ai-llm

Skill1 starsChanged 3 months agoMITView on GitHub

This is a thin human-facing router. Use skill.yaml as the source of truth for exact commands, inputs, outputs, and workflow chaining.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Start with surface-discovery unless prior evidence points to a more specific workflow.

    line 16
  2. Very specific:

    Follow each workflow next mapping in skill.yaml after reviewing generated findings.

    line 17
  3. Distinctive:

    If a workflow has no script reference, treat it as a manual or tool-native workflow and use the closest phase runbook when available.

    line 18

≈ “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

Ai Llm

Overview

AI/LLM Security Testing — prompt injection, tool abuse, output handling, RAG poisoning, MCP server attacks, training data extraction

This is a thin human-facing router. Use skill.yaml as the source of truth for exact commands, inputs, outputs, and workflow chaining.

Quick Reference

  • Skill: ai-llm
  • Severity range: info, low, medium, high, critical
  • Required tools: curl, jq, python3, httpx
  • Expected input files: live_hosts.txt, llm_api_paths.txt, mcp_endpoints.txt
  • Scope check: confirm authorization before running intrusive or authenticated testing.

Workflow Selection

  • Start with surface-discovery unless prior evidence points to a more specific workflow.
  • Follow each workflow next mapping in skill.yaml after reviewing generated findings.
  • If a workflow has no script reference, treat it as a manual or tool-native workflow and use the closest phase runbook when available.
  • Runbooks: use runbooks/ and select the closest phase runbook when workflow names do not map 1:1.

Available Workflows

WorkflowPurposeScript pathsPrimary outputsEvidence
surface-discoveryDiscover LLM endpoints, MCP servers, and chatbot surfaces.claude/skills/ai-llm/scripts/llm_surface_mapper.py$OUTDIR/ai-llm/surface_findings.jsonl<br>$OUTDIR/ai-llm/surface.json<br>$OUTDIR/ai-llm/mcp_endpoints.txt$OUTDIR/ai-llm/surface/evidence/
prompt-injectionTest direct and indirect prompt injection, jailbreak attempts.claude/skills/ai-llm/scripts/prompt_injection_runner.py$OUTDIR/ai-llm/prompt_injection_findings.jsonl$OUTDIR/ai-llm/prompt_injection/evidence/
tool-abuseTest excessive agency, tool/function calling abuse, plugin exploitation.claude/skills/ai-llm/scripts/tool_abuse_tester.py$OUTDIR/ai-llm/tool_abuse_findings.jsonl$OUTDIR/ai-llm/tool_abuse/evidence/
output-handlingTest if LLM output is rendered unsanitized (XSS, SSRF via injection).claude/skills/ai-llm/scripts/output_handling_tester.py$OUTDIR/ai-llm/output_handling_findings.jsonl$OUTDIR/ai-llm/output_handling/evidence/
rag-poisonTest RAG pipeline poisoning via document upload or URL injection.claude/skills/ai-llm/scripts/rag_poison_tester.py$OUTDIR/ai-llm/rag_poison_findings.jsonl$OUTDIR/ai-llm/rag_poison/evidence/
mcp-abuseTest MCP server for tool discovery, auth bypass, argument injection.claude/skills/ai-llm/scripts/mcp_fuzzer.py$OUTDIR/ai-llm/mcp_findings.jsonl$OUTDIR/ai-llm/mcp/evidence/
data-extractionExtract system prompts, training data, PII from LLM.claude/skills/ai-llm/scripts/data_extraction.py$OUTDIR/ai-llm/data_extraction_findings.jsonl$OUTDIR/ai-llm/data_extraction/evidence/

Evidence Required

  • Save raw request and response data for each confirmed finding.
  • Include timestamps, affected target, exact workflow name, tool versions, and reproduction steps.
  • Store screenshots or terminal captures in the workflow evidence directory when the workflow defines one.
  • Evidence templates from skill.yaml:
  • llm_response_dump: Full LLM response body saved as JSON
  • baseline_comparison: Diff between baseline and injected response
  • curl_poc: Reproducible curl command with payload
  • screenshot: Screenshot of chat widget showing injected output

References

  • Source of truth: skill.yaml
  • Runbooks: runbooks/
  • OWASP WSTG: WSTG-INPV-14, WSTG-CLNT-01

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.