The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository, such as feature briefs and software designs, which creates a surface for indirect prompt injection. Malicious instructions within these documents could be followed by the agent during the specification process.
Ingestion points: Local project files, including feature briefs (../../in-progress/feature-brief/SKILL.md), authoritative designs, and repository configuration artifacts.
Boundary markers: The instructions do not define specific delimiters or isolation techniques to prevent the agent from interpreting content within ingested files as instructions.
Capability inventory: The skill grants the agent the ability to inspect the repository (Step 1), launch executable targets (Step 3), and run project-defined validation commands (Validation section).
Sanitization: No sanitization or filtering logic is specified for the content retrieved from the repository files.
[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands and launch processes defined in the project's executable configuration to verify implementation feasibility and run validation tests (Step 3 and Validation section). While this is part of the intended software development workflow, it relies on the integrity of the project's local configuration.

implementation-spec