The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository that could influence autonomous behavior. * Ingestion points: The skill reads README.md, CONTRIBUTING.md, CONCEPTS.md, and Architecture Decision Records (ADRs) to ground its understanding of the project. It also extracts tracker URLs from the repository to identify external sinks for filing findings. * Boundary markers: No explicit delimiters or instructions are specified to prevent the agent from following commands potentially embedded in these documentation files. * Capability inventory: The skill is capable of executing shell commands (acceptance tests), modifying the file system, and performing network operations via git and tracker APIs. * Sanitization: There is no mention of sanitizing tracker URLs or validating the integrity of instructions retrieved from the repository files.
[COMMAND_EXECUTION]: The skill performs shell command execution during its autonomous phases. * Evidence: In Phase 5 (Verify), the skill autonomously executes acceptance commands defined in the Slice Contract. * Evidence: The skill executes python3 _shared/scripts/discover_runners.py to detect available runner environments.

ship