bank
Skill0 starsChanged 3 months agoMITView on GitHub
Bank what just happened into durable memory: lessons with what they cost and returned, decisions with their why, call debriefs routed into the vault. Run when the user says "what did we learn", "log this", "bank that", "debrief this call", pastes a transcript or meeting notes, or when anything meaningful just finished (a launch, a flop, a decision, a conversation). An event that ends only in chat history is an event the business forgets.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- jared833/claude-code-hookstheir own version, first seen Oct 21, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- jared833/claude-code-hooks★ 2same name
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. About 0 are common, 2 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 29
Before any step, read references/CONVENTIONS-core.md.
- Very specific: line 35
Run the shared contract's Universal preflight.
- Very specific: line 63
Run the shared durable claim admission gate on cost, result, and every sentence in the lesson.
- Distinctive: line 74
Ask before you assert: never present a quantity, duration, sample size, date, or causal motive in prose before asking the founder for it or …
- Distinctive: line 120
When a queue entry triggered the work, mark that same ID done with the learning/decision paths only after those files and their index links …
- Very specific: line 127
Run scripts/graph-audit on every touched durable file.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
bank: the business gets smarter or it just gets older
Week 20 beats week 1 only if what happened in between got banked. Bank writes the memory: lessons that carry their price tag, decisions that carry their why, conversations that update what the business knows about its people. Everything lands where recall will find it.
something happened
|
[ mode ] a lesson ──────────> LEARNING (cost + result attached)
| a hard call ───────> DECISION (3-test bar, 1-3 sentences)
| a call/meeting ────> DEBRIEF (extract + route, the headline mode)
|
[ dedupe ] similar note exists? UPDATE it, never clone it
|
[ file ] wiki/ + index + log found later, or worthless nowBefore any step, read references/CONVENTIONS-core.md. This resolves from the
installed skill directory to the plugin's shared contract, independent of the founder's
working folder. Its state, filing, decision, and dedupe rules govern this workflow.
Step 0: What kind of memory is this
Run the shared contract's Universal preflight. A Fresh, Partial, or Wrong-folder result stops and runs the co-founder-setup skill before anything else (re-sync for Partial), without creating a note or log row. On
Ready, read the charter's path map. Then classify the event into three shapes; one event can
produce several. Also read pending queue.md entries addressed to bank; review entries
point to the reviewed initiative and requested lesson rather than duplicating either.
- Learning: something was tried and taught something. Goes to
wiki/learnings/. - Decision: a call passing the shared contract's 3-test bar (hard to reverse,
surprising without context, a real tradeoff). Goes to
wiki/decisions/as its own page: base-five frontmatter (type: decision), 1-3 sentences of what and why, dated. Fails a test: it is not decision-log material, and logging it anyway is how decision logs die of noise. Awiki/decisions/**page is a T1 consent surface (shared contract, Immutable consent receipts): create it only with the founder's[receipt: raw/inbox/<file>#<exact words>]for the exact call, cited in the page. Absent a founder yes, write no decision page; a decision the founder never made is not banked as one. - Debrief: a conversation happened (customer, partner, investor, prospect). STOP and read
references/debrief-extraction.mdbefore touching it: the eight extraction sections and their routing are the contract, and a debrief filed as one blob is a debrief lost.
Done when: the shapes present in this event are named.
Step 1: Write the learning (when there is one)
The note follows the vault frontmatter with the money fields:
cost:what it took. Dollars when real, time always counts ("2 hrs", "3 wks").result:what happened, in numbers where numbers exist ("reply rate 4% -> 9% on 50 sends").status:honest confidence:highonly when the result would likely repeat.
Run the shared durable claim admission gate on cost, result, and every sentence in the lesson.
Reopen and quote the actual tracker or log row; a file path with blank totals is not a receipt.
Recompute calculated values from named inputs. When effort or outcome exists only in the founder's
recollection, write Founder-stated: unverified; when nobody supplied it, use not recorded
rather than manufacturing a plausible duration. Recheck original sources before carrying a claim
from review, content, or an older learning.
The shared parameter gate applies before filling any tidy-looking blank. Unknown effort, cost,
sample size, or result stays not recorded or founder-stated/unverified; the note shape never
justifies guessing a value.
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.
The writing contract: what failed gets written as a failure, in plain words. A flop reported
as "mixed results" poisons every future recall that touches it. History stays honest per the
shared dedupe rule: when today's evidence reverses yesterday's note, the note becomes the
current truth with the old claim preserved as superseded (supersedes: link) and why. One
canonical truth, with its history visible.
Done when: the note exists with cost, result, confidence, and its sources.
Step 2: Dedupe before filing
The shared dedupe rule, applied at EVERY destination this event
touches (learnings, decisions, people, the initiative), not just learnings: search
index.md, the fixed hub/<section>.md, and the target folder with the environment's file search. Found
substantially the same ground: UPDATE it. Fold the new evidence in, adjust the confidence,
bump updated:, note the supersession inside. Near-duplicates are how a vault rots into a
junk drawer where recall finds three half-truths instead of one truth.
Genuinely new ground: new note, base-five frontmatter whatever its type (person pages:
type: person, status: stated).
Done when: exactly one note carries this lesson, and it is the truest current version.
Step 3: The findability check
Answer concretely: which recall query, next month, should surface this note, and would it?
Fix what fails the check: the title says the insight plainly (a title like "notes from
Tuesday" fails), the tags carry the topic words a founder would actually use, the fixed hub
links it, index.md lists it with a one-line summary.
Done when: the note is reachable through index, its fixed hub, and a plausible search.
Step 4: Close the loops
- The event belongs to an initiative: add the dated line to that file's Log. The state flip
has a hard guard:
state: bankedONLY when the initiative currently readsstate: reviewedAND this run has filed or updated at least one learning page and linked its exact path from the initiative. A decision-only or debrief-only run cannot close the arc. Lessons from any other stage (a gauntlet DIES, a mid-execution surprise) link from the initiative's Log with the state untouched; banking a lesson never advances a lifecycle the work has not earned. A legalreviewed → bankedwrite also moves the initiative to the matching hub bucket and updates its index and baton projections before close. - Append the vault log row:
## [date] learning | <title>(ordecision/debrief). - When a queue entry triggered the work, mark that same ID done with the learning/decision paths only after those files and their index links are durable.
- Close in chat with the gist in two lines plus the file path. The note is the deliverable; the chat is just the receipt.
Mandatory filing gate
- Run
scripts/graph-auditon every touched durable file. - Fix every violation and rerun it. Learning, decision, lifecycle, log, and queue writes remain incomplete until it exits zero.
Done when: initiative, log, index, and chat all point at the same banked memory.
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.

