The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external codebase content into the agent's context through semantic search results, which is a known attack surface for indirect prompt injection.
Ingestion points: Code content retrieved via codecompass search:semantic and Weaviate CodeContext or AtlasCode collections.
Boundary markers: The guide does not provide specific instructions, delimiters, or system prompts to warn the agent about potentially malicious instructions embedded in code comments or logic within the retrieved snippets.
Capability inventory: The skill utilizes the codecompass CLI for searching and indexing and curl for service checks.
Sanitization: No sanitization, filtering, or validation of the retrieved code content is mentioned.
[COMMAND_EXECUTION]: The execution guide utilizes standard CLI tools to manage indexing and perform searches.
Evidence: Use of codecompass batch:index, codecompass search:semantic, and curl http://localhost:8081/v1/schema. These operations target local services on localhost and are consistent with the skill's stated purpose for local code discovery.

semantic-search