The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow in SKILL.md instructs the agent to execute a specific binary tool named convex-scan located at /home/bjorn/.codex/skill-support/bin/convex-scan. This is a vendor-provided tool used for scanning the repository backend.- [EXTERNAL_DOWNLOADS]: The agents/openai.yaml configuration defines dependencies on external Model Context Protocol (MCP) tools, specifically context7 for API documentation and exa for web-based research.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from the repository's documentation.
Ingestion points: The skill reads AGENTS.md, README.md, and architecture documentation in SKILL.md (Step 1) and scripts/build_feature_seed.py (line 20).
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present when reading repository files.
Capability inventory: The skill has the capability to execute shell commands (convex-scan) and write files to the disk (scripts/build_feature_seed.py).
Sanitization: No sanitization or validation logic is implemented for the data retrieved from repository files.

convex-feature-spec