from joelclaw50
Fan-out retrieval pattern that searches all memory sources when a vague reference to past context appears, preventing guesswork and ensuring accurate context be
Implements a fan-out recall pattern: when a user makes a vague or referential request ("from earlier", "remember when", "that thing with"), the skill searches multiple memory sources in parallel (daily logs, curated memory, session transcripts, vault notes, system logs, processed media, Redis state) to find the referenced context. It then synthesizes results into a concise summary for the agent or user.
Recall is a fan-out retrieval skill that searches multiple memory sources (daily logs, curated memory, session transcripts, vault notes, Redis) when a user references past context vaguely. It has no bundled scripts — all logic lives in SKILL.md as shell command patterns. Well-structured trigger detection and anti-patterns section, but entirely prescriptive with no executable code.
No security concerns — no remote fetch, no credentials, no destructive commands. Shell commands use timeout which is good. Portability is limited by macOS-specific date flags and joelclaw-specific paths. The skill is useful as a pattern but very niche (personal to Joel's setup) and would need adaptation for broader use.
