The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/search.sh detects the user's platform and executes a corresponding pre-compiled binary (e.g., ch-search-darwin-arm64). Since the source code for these binaries is not provided, their internal operations, such as file system access beyond the stated scope or network communications, cannot be verified.
[PROMPT_INJECTION]: The skill exposes the agent to indirect prompt injection by retrieving and processing historical conversation data.
Ingestion points: The skill indexes conversation history located at ~/.claude/projects/.
Boundary markers: The prompt instructions do not specify any delimiters or safety markers to help the agent distinguish between search results and instructions.
Capability inventory: The skill allows for the execution of arbitrary platform-specific binaries provided in the bin/ directory.
Sanitization: There is no evidence of content sanitization or instruction-stripping performed on the conversation snippets retrieved from the database.

conversation-search