The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process historical data ("memories") which may contain untrusted content from previous sessions.
Ingestion points: Reads markdown files from .memsearch/memory/ and conversation transcripts via transcript.py.
Boundary markers: Absent. The instructions do not define delimiters or warnings for the agent when processing retrieved memory chunks.
Capability inventory: The skill has access to the Bash tool and can execute shell commands, memsearch CLI, and Python scripts.
Sanitization: Absent. Content retrieved from memory is summarized directly for the user context.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file uses the dynamic injection syntax (!command) to determine the project collection name at skill load time.
Evidence: Collection: !bash -c 'if [ -n "${MEMSEARCH_DIR:-}" ]; then ...'`
Details: This command performs environment discovery (checking MEMSEARCH_DIR, CLAUDE_PLUGIN_ROOT, and git) to resolve pathing for the memory database. The operation is benign and supports standard project-specific tooling functionality.
[EXTERNAL_DOWNLOADS]: The skill instructions suggest using uvx to execute the memsearch package if it is not already present in the environment.
Evidence: If memsearch is not found, try uvx memsearch instead.
Details: uvx is a tool for running packages from the Python Package Index (PyPI). In this context, it is used to ensure the core utility for the skill is available.
[COMMAND_EXECUTION]: The skill directs the agent to execute various shell commands to explore and retrieve data.
Evidence: Instructions include usage of grep, ls, cat, and git rev-parse to navigate the .memsearch directory and identify recent logs or session headings.

memory-recall