The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from various files within the local repository to generate instructions, which could contain malicious content if the codebase is untrusted.\n
Ingestion points: The skill reads package.json, .ds-ops-config.yml, CONTRIBUTING.md, CHANGELOG.md, and existing instruction files like AGENTS.md or CLAUDE.md to identify rules and repository structure.\n
Boundary markers: The skill relies on a 'provenance rule' which requires rules to be sourced from specific files or URLs, but it lacks explicit boundary delimiters or instructions to ignore embedded prompts in the ingested data.\n
Capability inventory: The skill possesses the capability to write files to the repository root and use shell commands to read/search the filesystem.\n
Sanitization: Content from files is directly summarized and interpolated into the generated AGENTS.md without explicit sanitization beyond the requirement for sourcing.\n- [COMMAND_EXECUTION]: The skill uses shell commands to perform repository discovery and file inspection.\n
Evidence: The skill logic involves searching for system parts and reading configuration files, utilizing tools like find, ls, and cat as described in its inventory steps.

agent-instructions