skill-improver

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from repository files and session history to generate improvements.
  • Ingestion points: The skill reads repository artifacts, including scripts, CI configurations, and other skill definitions (SKILL.md, Step 1 & 2).
  • Boundary markers: The instructions lack clear boundary markers or warnings to the agent to ignore instructions embedded within the files being analyzed.
  • Capability inventory: The skill has the capability to modify files and execute shell commands for linting and testing (SKILL.md, Step 3 & 4).
  • Sanitization: There is no mechanism to sanitize or validate the content of analyzed files before they are used to influence the agent's logic or generated code.
  • [COMMAND_EXECUTION]: The skill requires the execution of shell commands as part of its standard verification process.
  • Evidence: Step 4 in SKILL.md instructs the agent to "Run format/lint commands" and perform a "smoke test" on modified scripts, which involves executing code that might have been maliciously altered via the improvement process.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 02:13 PM
Security Audit — agent-trust-hub — skill-improver