skills/gharib89/skills/update-skills/Gen Agent Trust Hub

update-skills

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill fetches commit subjects and changelog entries from external GitHub repositories and includes them in the generated Pull Request body. This creates an attack surface where a malicious actor could inject instructions into the PR description to influence agents reading it.
  • Ingestion points: heads.sh and plan.sh fetch commit subjects via the GitHub API and read CHANGELOG.md files from the project worktree.
  • Boundary markers: Content is placed inside markdown <details> and <summary> tags, but lacks explicit "ignore embedded instructions" warnings.
  • Capability inventory: The skill utilizes open-pr.sh, file-issue.sh, and npx skills add to modify the repository and external state.
  • Sanitization: No sanitization, escaping, or filtering of the ingested commit subjects or changelog text is performed before interpolation into the PR body.
  • [DYNAMIC_EXECUTION]: The skill dynamically constructs and executes commands using eval to install or update skill packages.
  • Evidence: SKILL.md uses eval to run npx skills@1.8.1 add Gharib89/skills $flags ... where $flags is built from JSON lock files.
  • Mitigation: The skill uses the @sh filter within jq to properly escape strings, reducing the risk of shell injection.
  • [EXTERNAL_DOWNLOADS]: The skill downloads and installs packages and skill definitions from GitHub.
  • Evidence: Usage of npx skills@1.8.1 add and curl requests to api.github.com to fetch repository data.
  • Context: These operations target the author's own infrastructure or well-known services (GitHub), which is consistent with the skill's stated purpose.
  • [COMMAND_EXECUTION]: The skill executes various shell commands and scripts, including complex pipelines using bash -c for file searching and term replacement.
  • Evidence: Multiple instances of bash -c are used in SKILL.md for task automation during the skill refresh process.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 11:13 AM