recall
Skill0 starsChanged 3 months agoMITView on GitHub
Check what the business already knows before deciding anything. Run when the user says "what do we know about...", "have we tried this before", "check the vault", "what did we learn about X", and AUTOMATICALLY before any gauntlet, plan, offer, or big decision, per the charter's recall-first ritual. Answering a business question from model memory when the vault holds the founder's real history is the failure this skill exists to prevent.
153 other projects use a version of this.
Adopted by
Most recently taken up by
- TencentBlueKing/bk-litetheir own version, first seen Nov 16, 2026
- TechNickAI/hermes-skillstheir own version, first seen Nov 16, 2026
- GoBeromsu/craft-skillstheir own version, first seen Nov 16, 2026
- gfunc/skillstheir own version, first seen Nov 16, 2026
- DeepExtrema/Tesseracttheir own version, first seen Nov 16, 2026
- shiwenbin1617/pstacktheir own version, first seen Nov 16, 2026
Used with
Projects with it also use addskill, apply-firmware and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
ryanportfolioSkilladdskill
Create, import, update, or install repo or personal skills: runtime ownership, resources, discovery checks.
Also in 6 projects that use this
ryanportfolioSkillapply-firmware
Bring current folder (plain folder, existing repo, old or partial firmware copy) up to full Harness Firmware in place. Mirror external repo → adopt-repo.
Also in 6 projects that use this
- +10 more from ryanportfolio
Other versions153 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 153
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- backnotprop/pstack★ 4.2ksame name
- parcadei/Continuous-Claude★ 3.9ksame name
- parcadei/Continuous-Claude-v2★ 3.9ksame name
- parcadei/Continuous-Claude-v3★ 3.9ksame name
- indranilbanerjee/digital-marketing-pro★ 934same name
- ArtemXTech/personal-os-skills★ 545same name
- entireio/skills★ 240same name
- termdock/termdock-issues★ 160same name
143 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 1 is distinctive, and 5 are very specific to this project.
0
Common
Most of the 140 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 31
Read references/CONVENTIONS-core.md and run its Universal preflight.
- Very specific: line 39
Read index.md first: it is the catalog, and it says where to drill.
- Very specific: line 48
Before saying anything shaped like "we have no note on that": list the relevant wiki/ folders and search the question's key words (and …
- Very specific: line 71
Read canonical lifecycle fields before using status words.
- Distinctive: line 77
Ask before you assert: never present a quantity, duration, sample size, date, or causal motive in prose before asking the founder for it or …
- Very specific: line 103
When recall writes a queue item, run scripts/graph-audit queue.md.
“Common” is measured: kept by at least 6 in 10 of the 140 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
recall: the vault answers first
The most expensive sentence an AI can say to a founder is "we have no notes on that" when the notes exist. Recall reads before it speaks, cites what it finds by exact file, and admits plainly when the vault is genuinely empty.
the question
|
[ index ] read index.md first, drill into pages
|
[ verify ] list + grep BEFORE any "nothing on record"
|
[ answer ] cited by file, confidence surfaced
|
[ bank ] a real synthesis goes back into the vaultStep 0: Preflight
Read references/CONVENTIONS-core.md and run its Universal preflight. A Fresh,
Partial, or Wrong-folder result stops and runs the co-founder-setup skill before anything else (re-sync for Partial);
an absent vault cannot be searched honestly. On Ready, read the charter's path map.
Done when: the founder system is Ready and the catalog path is known.
Step 1: Retrieve
Read index.md first: it is the catalog, and it says where to drill. Then read the pages
that plausibly bear on the question, plus the matching fixed hub/<section>.md. Search wide before answering
narrow: the useful lesson is often filed under the channel, the person, or the initiative
rather than the topic word.
Done when: every plausibly relevant page has been looked at, not guessed at.
Step 2: The absence guard
Before saying anything shaped like "we have no note on that": list the relevant wiki/
folders and search the question's key words (and their obvious synonyms) across the vault
with whatever the environment has (rg or grep where present, the harness's search or
explicit listing plus reading otherwise, per CONVENTIONS). Absence gets VERIFIED, never
remembered: answering from memory instead of searching is the single most common way systems
like this fail. Genuinely empty: say exactly that, in one line, and offer the research skill
to fill it.
Any durable absence statement or queue item records the shared verified-absence fields: paths or corpus, terms and synonyms, date, tool, and result count. Phrase only the scoped result. It can prove what is not recorded in the vault, never what the business or world has never done.
Done when: presence and absence are both grounded in an actual look.
Step 3: Answer with receipts
- Cite the exact file for every claim:
wiki/learnings/warm-intro-beats-cold.md, not "our notes." Run the shared durable claim admission gate: reopen the file, verify its current lifecycle and confidence, and quote the exact supporting passage. Trace derived claims to their original inputs. A page that calls a draft approved, or mentions a topic without the claimed evidence, does not pass merely because recall found it. - Surface confidence: a
statedclaim is repeated as a quote with its source, never as a truth. Alownote gets its doubt said out loud. - Read canonical lifecycle fields before using status words. An offer is approved only at
offer_state: approvedoractive; publication exists only atcontent_state: publishedwithpublished_at. Contradictory index or hub prose is reported as drift, not repeated as truth. - Cross-file analogies and patterns pass the shared parameter gate. When no admitted receipt states
the synthesis, label it
Unverified synthesis:and offer the one founder question that could confirm it; rhetoric never upgrades itself into remembered fact. - Ask before you assert: never present a quantity, duration, sample size, date, or causal motive in prose before asking the founder for it or citing a receipt for it.
- Notes past their
review_afterdate get flagged as possibly stale. - Feeding another skill (gauntlet, plan, offer): deliver a task-scoped mini-brief, the 5-10 lines that bear on THIS decision, never a vault dump.
- Exclude every
type: contentpage from business-fact receipts. Drafts are proposals, and approved/published pieces are publication history, not the source of truth. Usewiki/business/offer.mdfor the current offer andwiki/business/metrics.mdfor current numbers. - This is the internal leg of the receipts law; when the question also needs external truth, say so and name research as the other leg.
Done when: the founder can click through to every fact in the answer.
Step 4: Bank real syntheses
A recall that produced genuine new synthesis (a comparison, a pattern across notes, a
connection nobody had written down) is itself a finding. Append one pending entry to
queue.md addressed from recall to steward. The request carries the exact synthesis and the
source_paths list names every note it came from; recall reads, it never creates the wiki page.
Narrate the queue ID in one line. A synthesis that evaporates into chat history has to be paid
for again next month. Routine lookups do not get banked; the test is whether the answer contains
something no single existing page says.
Mandatory filing gate
- When recall writes a queue item, run
scripts/graph-audit queue.md. - Fix every violation and rerun it. The queue item is not durable and no queue ID is narrated until it exits zero. A read-only lookup writes nothing and has no filing gate.
Done when: anything newly true has a durable pending queue ID, and everything else is cited.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

