spec-charter
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted repository metadata to generate project specifications and agent instruction files, which could allow malicious repository content to influence agent behavior.
- Ingestion points: The skill reads README.md, CHANGELOG.md, CLAUDE.md, AGENTS.md, git commit history, and the text of open issues, epics, and pull requests as described in create.md and alignment.md.
- Boundary markers: Generated content in harness files is delimited by spec-charter-projection markers.
- Capability inventory: The skill writes files to the spec/ directory, modifies agent harness files (AGENTS.md, CLAUDE.md), and executes shell commands for git and local script execution.
- Sanitization: There is no evidence of sanitization or filtering of external content (such as issue titles) before it is interpolated into the proposed charter objectives.
- [COMMAND_EXECUTION]: The skill executes git commands and is designed to run local scripts if they exist within the user's repository.
- Evidence: The reassessment logic in reassess.md specifically looks for and executes helper scripts such as capabilities-doctor.js and component-lint.js when present in the target repository's environment.
Audit Metadata