setup-matt-pocock-skills
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using standard development tools, including git, the GitHub CLI, and the GitLab CLI. These commands are used to discover repository metadata (e.g.,
git remote -v) and interact with issue trackers (e.g.,gh issue list). - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, including GitHub/GitLab issue bodies and comments, as well as local markdown files in the
.scratch/directory. This creates a surface for indirect prompt injection where malicious instructions in issue content could attempt to influence the agent's configuration tasks. - Ingestion points: Untrusted data is ingested via commands like
gh issue viewandglab issue view, and by reading files from.scratch/. - Boundary markers: The skill does not define explicit boundary markers or instructions to ignore embedded directives when processing the external content.
- Capability inventory: The skill has the capability to execute shell commands (
git,gh,glab) and write configuration files to the repository (e.g.,docs/agents/issue-tracker.md,CLAUDE.md). - Sanitization: While the instructions recommend using shell heredocs for multi-line CLI inputs to prevent simple command injection, there is no logic to sanitize or filter the content for prompt injection attacks.
Audit Metadata