How do I install the Recall skill?

Run npx skillmds@latest add mk668a/recall in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Recall safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Recall?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Recall free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Recall?

mk668a (@mk668a) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Recall by mk668a — Agent Skill

Name: Recall
Author: mk668a

Recall

Use when the user wants to apply lessons from past engagements to the current one (e.g. "what did we learn about pilots like this", "any priors for findings reports", "how have we handled this risk before"). Reads the anonymized corpus and surfaces relevant priors, weighted by how often they recurred — without exposing any past client.

mk668a 944a2f3 1.7 KB Updated Sep 17, 2026

File contents

fde-recall — pull relevant priors from the corpus

The read side of compounding. Lets the current engagement benefit from every past one, while the corpus stays client-anonymous.

Steps

Take the user's question or the current deliverable <type> as the query.
Read .fde/corpus/<type>.yml (or all of .fde/corpus/*.yml for a broad question). Every entry is already anonymized (<CLIENT>, $<REDACTED>, <N>%) and carries sightings and paved.
Surface the 3-5 most relevant priors as patterns, not verbatim dumps, and weight them by tier: a paved entry is hardened practice ("this held across 3+ engagements"), a ×2 is recurring, a ×1 candidate is one engagement's opinion — say which is which.
Hand off: suggest /fde:draft to fold these into the current deliverable (where paved entries pre-fill and candidates become [candidate ×n — confirm] suggestions).

Hard rules

Priors come from .fde/corpus/ only. Never read another engagement's engagements/<other-client>/ to answer a recall question — the corpus is the only legitimate cross-engagement source, because it is the only layer the anonymization gate guards.
If the corpus is empty (early days), say so honestly — there are no priors yet, and the first /fde:promote is how that changes.

Frequently asked questions