source-command-cowork-veille

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads from a hardcoded absolute local path at /Users/florianbruniaux/Sites/perso/Codex-cowork-guide/CHANGELOG.md. While this path is consistent with the author's environment, absolute paths increase the risk of exposing local file structures.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from web searches and external websites and writes it into a local project file, which could execute malicious instructions embedded in the source data.
  • Ingestion points: External content fetched via perplexity_search and WebFetch (SKILL.md).
  • Boundary markers: Absent; the skill does not use specific delimiters or instructions to ignore embedded commands in the fetched content.
  • Capability inventory: The agent has the capability to write to the local filesystem to update the CHANGELOG.md file.
  • Sanitization: No sanitization or validation of the fetched external content is performed before writing to the local file.
  • [COMMAND_EXECUTION]: The skill modifies the local filesystem by reading and appending content to a project file based on external input.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 03:29 AM
Security Audit — agent-trust-hub — source-command-cowork-veille