The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The from-trd feature ingests external binding specifications (TRD/SoW) to generate project documentation and planning shapes.
Ingestion points: The skill reads files from user-defined paths specified in .rigorous/trd-config.yml (e.g., docs/spec.md).
Boundary markers: The instructions do not define strict boundary markers or instructions to ignore potential prompt injections embedded within the external specifications.
Capability inventory: The skill has broad capabilities including file system modification, script generation, and subagent orchestration which could be influenced by malicious content in the spec.
Sanitization: Content from the specification is interpolated into documentation and planning files without explicit sanitization or escaping.
[DYNAMIC_EXECUTION]: The skill generates and manages executable scripts and project-level hooks from local templates.
Evidence: The templates/ folder contains boilerplate for .mjs and .sh scripts that are concretized and written to the project's tools/scripts/ directory during setup.
Impact: Installs platform-level PreToolUse hooks in .claude/settings.json which intercept shell commands to enforce project-specific policies.
[PERSISTENCE]: The skill provides a mechanism to install standalone command shortcuts by modifying the agent's environment.
Evidence: The pin command programmatically creates new skill definition files in the ~/.claude/skills/ directory.
[COMMAND_EXECUTION]: Routinely executes various CLI tools to gather project metadata and perform technical audits.
Evidence: The skill invokes commands such as git diff, gh pr list, npm audit, trivy, and semgrep across its synchronization and auditing workflows.

rigorous