The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and analyzing untrusted project data (source code, repository metadata, performance logs) while having the permission to modify the code base, which creates a surface for indirect prompt injection.
Ingestion points: The skill reads repository-level agent instructions, source code files, repository commits, and running binary/version data (SKILL.md, Workflow Step 1).
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious content inside the analyzed project files.
Capability inventory: The skill is authorized to implement code repairs, write Architecture Decision Records (ADR), and instrument code paths (SKILL.md, Workflow Steps 8 and 10).
Sanitization: The skill does not mention specific sanitization or escaping of the ingested project data before it is processed by the agent.
[COMMAND_EXECUTION]: The diagnostic workflow (Workflow Step 1) involves the agent executing commands to collect runtime evidence, such as measuring CPU/memory usage, query counts, and subprocess depth. Users should monitor the specific commands the agent chooses to run for these metrics.
[EXTERNAL_DOWNLOADS]: The skill requires the PyYAML Python package for its local validation script (scripts/validate_skill.py). This is a well-known and standard package in the Python ecosystem.

lov-app-professional-design