release
Warn
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: MEDIUMPERSISTENCECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PERSISTENCE]: The
scripts/install-hooks.shscript modifies the local.git/hooksdirectory to install apre-pushhook. This ensures that the scriptscripts/pre-pushis executed automatically whenever agit pushcommand is issued. While this is a documented feature of the skill, the installation is performed silently byscripts/release-context.sh(redirecting output to/dev/null), which can obscure the modification of the repository's configuration.\n- [COMMAND_EXECUTION]: The skill instructions and scripts execute various shell commands and local scripts, includinggit,jq,awk,gh, and several.shfiles located within thescripts/directory. It is important to note that the content ofscripts/release.sh,scripts/extract-changelog.sh, andscripts/pre-pushwas not provided in the analyzed file set.\n- [INDIRECT_PROMPT_INJECTION]: The skill reads and processes untrusted data from the repository, including commit messages andCHANGELOG.mdcontent, to generate release highlights and changelogs.\n - Ingestion points:
scripts/release-context.shingests output fromgit logand the contents ofCHANGELOG.md.\n - Boundary markers: The provided scripts do not use specific delimiters or instructions to prevent the agent from following potential commands embedded within commit messages or changelog entries.\n
- Capability inventory: The skill has the ability to write to files (
CHANGELOG.md,package.json), execute local shell scripts, and perform network operations viagit pushand the GitHub CLI (gh).\n - Sanitization: There is no evidence of sanitization or filtering of the ingested git metadata before it is presented to the agent.
Audit Metadata