from ai-toolkit150
Search past coding sessions and observations using natural-language queries to retrieve decisions, notes, and context.
Searches a local SQLite FTS5 memory database to find past coding observations, decisions, and session context. Returns a concise summary view first and can show full observation details on request. Useful for recalling prior decisions, debugging notes, and project history across sessions.
Use when you need to find prior notes or decisions from previous coding sessions (e.g., migrations, design choices, troubleshooting steps). Trigger when asking for historical context like "what did we decide about X" or "show past notes about Y".
Best suited for local/code-focused assistants that can run Bash/Python and access a user home directory (Copilot/Codex/Claude Code style agents).
mem-search is a query-only skill that searches a SQLite FTS5 index of past coding session observations. It has no scripts to run. The SKILL.md is well-structured with clear instructions, progressive disclosure (summary then detail), and useful gotchas. Main security concern is FTS5 query injection — the skill mentions escaping single quotes but doesn't fully sanitize FTS5 operator syntax, which can cause malformed MATCH expressions or unintended query logic.
No scripts present — static analysis only. The skill is functional and well-documented. The FTS5 injection risk is real but low-severity (query errors, not data exfiltration). Usefulness is moderate since it's tied to the softspark ecosystem.
