report
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes
gitcommands using thespawnSyncmethod inscripts/weekly.jsandsrc/weekly.tsto retrieve commit logs, project metadata, and user configuration. This is a primary function of the skill. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It gathers Git commit messages and sends them to an LLM to be 'optimized' into a reporting format based on instructions in
resources/prompt.txt. An attacker who can commit to a repository scanned by this skill could embed instructions to manipulate the report's content or the agent's behavior. - Ingestion points: Git commit messages are retrieved from local repositories in
scripts/weekly.js. - Boundary markers: Absent. The LLM prompt in
resources/prompt.txtdoes not use delimiters or specific safety instructions to distinguish between the report generation logic and the untrusted commit data. - Capability inventory: The skill executes local Git commands and writes output to
.jsonand.docxfiles on the local filesystem viascripts/weekly.jsandscripts/weekly_render.js. - Sanitization: Absent. Commit messages are passed to the LLM without escaping or filtering.
Audit Metadata