sales-request-skill
Pass
Audited by Gen Agent Trust Hub on Apr 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Uses shell commands including
find,cat, and the GitHub CLI (gh) to manage local files and interact with remote repositories. It also executes a Python snippet viasubprocess.runto launch a web browser with pre-filled submission data.- [DATA_EXFILTRATION]: Aggregates data from locallearnings.mdfiles across installed skills to share them with thesales-skills/salesrepository. This behavior is the primary function of the skill and includes documented steps for manual PII scrubbing and user confirmation before transmission to the vendor's own repository.- [DYNAMIC_EXECUTION]: Employs Python'ssubprocessmodule to programmatically open a specific browser profile for PR/Issue review, representing a form of runtime command execution.- [INDIRECT_PROMPT_INJECTION]: - Ingestion points: Reads content from
~/.claude/skills/*/references/learnings.mdfiles which are technically external sources. - Boundary markers: None implemented at the file read stage; the skill relies on subsequent manual review steps.
- Capability inventory: Can execute GitHub CLI commands (
gh pr create,gh issue create) and system commands via Python. - Sanitization: Includes a dedicated 'Step C3' for scrubbing personally identifiable information (PII) before any data is sent externally.
Audit Metadata