CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

auditing-mcp-servers-for-tool-poisoning

Audit MCP servers for tool poisoning, tool shadowing, rug pulls, SSRF, and unauthenticated exposure using Invariant Labs' mcp-scan for static/runtime scanning plus manual SSRF/auth checks and description pinning. Use before adding a new MCP server to an agent stack, when reviewing an internal MCP server, detecting rug pulls, or investigating an agent's unexpected tool-driven behavior.

Quality

84%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

Quality

Content

72%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is highly actionable with executable code and a clear 8-step sequence, but it underuses its own bundle files (no inline links to references/scripts) and lacks in-workflow validation checkpoints for a process that includes risky probing steps.

Suggestions

Add in-line validation checkpoints to the workflow (e.g., after the static scan, confirm no high-severity findings before proceeding to runtime probing) and a fix/retry loop for the SSRF and enumeration steps.

Link the existing bundle files from the body — point to references/api-reference.md for full CLI/SDK detail and references/standards.md for ATLAS/NIST/OWASP mappings — and move the duplicated MITRE/standards tables out of the body to remove inline duplication.

Reference scripts/agent.py in the Workflow so the provided audit helper is discoverable instead of only hand-rolling enumerate/ssrf probes inline.

DimensionReasoningScore

Conciseness

Mostly efficient with executable code and a tight structure, but the Overview re-explains tool-poisoning background and the threat/MITRE tables duplicate content in references/standards.md, adding minor padding.

4 / 5

Actionability

Provides copy-paste-ready executable bash (uvx mcp-scan, curl, ss) and complete Python (enumerate_tools.py, ssrf_probe.py) covering the common audit cases with specific commands and concrete examples.

5 / 5

Workflow Clarity

Eight steps are clearly numbered and sequenced, but the workflow has no in-line validation checkpoints or fix/retry loops, and the SSRF probing step is a risky operation without validation gating — capping per the rubric's batch/risky-operations rule.

3 / 5

Progressive Disclosure

Section structure is good, but existing bundle files (references/api-reference.md, references/standards.md, scripts/agent.py) are never linked or signaled from the body, and content that belongs in those references (API reference, ATLAS/standards tables) is duplicated inline rather than pointed to.

3 / 5

Total

15

/

20

Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is specific, complete, and clearly distinguishable, with strong concrete action coverage and explicit 'when' triggers. Trigger term coverage is strong but could add a few more synonyms to reach the top anchor.

DimensionReasoningScore

Specificity

Lists multiple concrete auditing actions — tool poisoning, shadowing, rug pulls, SSRF, unauthenticated exposure — plus specific methods (mcp-scan static/runtime scanning, manual SSRF/auth checks, description pinning), giving comprehensive coverage.

5 / 5

Completeness

Explicitly answers both 'what' (audits MCP servers for the listed threats via mcp-scan, manual checks, and pinning) and 'when' with concrete trigger phrases in a clear 'Use before/when/or'-style clause.

5 / 5

Trigger Term Quality

Includes natural trigger phrases a user would say ('adding a new MCP server to an agent stack', 'reviewing an internal MCP server', 'detecting rug pulls', 'investigating an agent's unexpected tool-driven behavior') with good keyword coverage, though a few synonyms/extensions are absent.

4 / 5

Distinctiveness Conflict Risk

Targets a clear niche (MCP server security auditing for tool poisoning) with distinct, specific triggers that are unlikely to fire for unrelated skills.

5 / 5

Total

19

/

20

Passed

Validation

93%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

15

/

16

Passed

Repository

mukul975/Anthropic-Cybersecurity-Skills

Path

skills/auditing-mcp-servers-for-tool-poisoning/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.