The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by scanning and processing arbitrary repository files.\n
Ingestion points: The skill uses read, glob, and bash commands to ingest content from root config files (e.g., package.json, Cargo.toml, requirements.txt), README.md, and the docs/ directory.\n
Boundary markers: There are no explicit markers or specific instructions provided to the agent to ignore potentially malicious embedded instructions within the ingested repository content.\n
Capability inventory: The skill utilizes shell execution (bash, bd) and file writing capabilities to create or refactor context files like CLAUDE.md.\n
Sanitization: Content extracted from repository files is incorporated into documentation without evident sanitization or validation mechanisms.\n- [COMMAND_EXECUTION]: The skill executes local shell commands to perform repository analysis and task management.\n
Evidence: Uses bash for tech stack analysis and executes bd prime to check for the presence of the BEADS project management system. It also suggests the use of the bashcheck tool for syntax validation.

organize-ai-context