The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses dynamic context injection (!bash) to automatically identify the repository root and derive the vector database collection name when the skill is first loaded.
[COMMAND_EXECUTION]: The instructions direct the agent to use the Bash tool for running the memsearch utility, performing filesystem searches with grep, and executing a local Python script for transcript processing.
[EXTERNAL_DOWNLOADS]: The skill recommends using the uvx tool to dynamically download and execute the memsearch package from the Python Package Index (PyPI) if it is not found on the local path.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes content from historical conversation logs and project notes that could potentially contain malicious instructions from prior sessions.
Ingestion points: Historical session files in .memsearch/memory/*.md and transcript JSONL files referenced in the search results.
Boundary markers: Absent; there are no specific instructions or delimiters provided to ensure the agent ignores embedded instructions within retrieved data.
Capability inventory: The skill uses the Bash tool for search operations, filesystem traversal, and local script execution as defined in SKILL.md.
Sanitization: Absent; retrieved memories are summarized and returned to the primary conversation context without prior validation or filtering.

memory-recall