housekeep-dead-code

Pass

Audited by Gen Agent Trust Hub on Sep 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill makes heavy use of npx to run utilities like knip, jscpd, and remove-unused-vars. Specifically, it identifies remove-unused-vars as a 'highly experimental' tool (version 0.0.x), which introduces a minor supply chain risk as it is executed directly from the npm registry without a pinned version in the command string.
  • [COMMAND_EXECUTION]: The skill instructions automate the execution of shell commands, including piping output from a linter directly into a code-modifying tool (npx eslint ... -f json | npx remove-unused-vars). While efficient, this automated modification of the filesystem relies on the integrity of the linter output.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from tool outputs and potentially acting upon it.
  • Ingestion points: Reads JSON output from knip --reporter json (into knip-files.json) and eslint -f json via pipes.
  • Boundary markers: The skill partially mitigates this by requiring an 'approved plan-dead-code.md' (human-in-the-loop) and explicitly warning not to 'delete from a raw tool dump'. However, no machine-level delimiters or sanitization are used in the shell pipes themselves.
  • Capability inventory: The skill uses knip --fix-type and remove-unused-vars which have capabilities to delete files and modify source code.
  • Sanitization: There is no mention of escaping or validating the JSON output from tools before it is processed by subsequent scripts or the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 13, 2026, 05:12 AM
Security Audit — agent-trust-hub — housekeep-dead-code