The Agent Skills Directory

[PERSISTENCE]: The install-hook script modifies the ~/.claude/hooks/sessionEnd lifecycle hook to automatically execute the indexing tool whenever a session concludes, ensuring its code runs periodically without user intervention.
[DATA_EXFILTRATION]: The summarizer.ts component reads user conversation logs from the local file system and transmits the content to the Anthropic API via the @anthropic-ai/claude-agent-sdk to generate summaries.
[EXTERNAL_DOWNLOADS]: The embeddings.ts file utilizes the @xenova/transformers library, which fetches the all-MiniLM-L6-v2 model from Hugging Face's official repository during the initial execution.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves summarizing and indexing conversation logs, which creates a potential surface for indirect prompt injection if the processed logs contain malicious instructions from untrusted sources.
[COMMAND_EXECUTION]: Several shell scripts (install-hook, index-conversations, search-conversations) perform file system modifications and execute TypeScript logic using npx tsx.

remembering-conversations