remember-this — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited remember-this (Agent Skill) and scored it 92/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 2 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 2 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
When the user explicitly tells you to remember something, save it to CogmemAi immediately. Don't wait until the end of the session, don't paraphrase loosely, and don't ask for confirmation. The user has done the hard work of identifying what matters; your job is to persist it correctly.
Save the moment you hear any of:
Also trigger when the user states a stable preference even without an explicit "remember" phrase. For example: "I always use Bun instead of npm" is a preference worth saving even though they didn't say "remember".
Use save_memory with:
content — restate the user's fact in one or two complete sentences. Include the why if they gave one. Concrete and self-contained.memory_type — usually preference for user preferences, decision for project decisions, identity for facts about the user, architecture for tech-stack facts.importance — usually 7-9 for explicit "remember this" requests. The user flagged it as important, respect that signal.scope — global if the fact applies across all projects (user preference, identity, tool choice). project if it's specific to the current codebase.tags — short labels so related memories cluster (e.g., ["coding-style"], ["deployment"], ["communication"]).save_memory({
content: "User has a strict no-em-dash rule. Never use the em dash character (—, U+2014) in any output: chat replies, code comments, docs, commit messages, marketing copy. Use a comma, parentheses, period and new sentence, colon, or semicolon instead.",
memory_type: "preference",
importance: 9,
scope: "global",
tags: ["writing-style", "punctuation"]
})save_memory({
content: "Deploy target for this project is Google Cloud Run, not AWS Lambda (decided 2026-05-13). Switch was driven by per-request billing simplicity and easier local testing with Docker.",
memory_type: "decision",
importance: 9,
scope: "project",
tags: ["deployment", "infrastructure"]
})save_memory({
content: "Always format git commit messages using Conventional Commits (feat:, fix:, chore:, docs:, refactor:, test:). Subject under 72 characters. Body wrapped at 72.",
memory_type: "preference",
importance: 8,
scope: "global",
tags: ["git", "commits"]
})update_memory instead so version history is preserved.Briefly confirm so the user knows it was captured. One short line. For example: "Saved as a global preference."
cogmemai-memory — full reference for memory management, types, and importance scoringsave-context — capture multi-fact context vs. a single explicit instructionsession-start — these saved memories surface again via get_project_context~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.