bank-map
Skill1 starsChanged 4 months agoMITView on GitHub
Regenerate bank/evidence_map.md, the drafter's derived index of the experience bank. Run after any edit to bank/experience_bank.md (new role, new contribution, updated facts). Reads the full bank, rebuilds the map from scratch.
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 50
When done, confirm the map's unit count matches the bank's (every contribution and project has exactly one angle-inventory entry) and stop.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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-map: regenerate the evidence map
bank/evidence_map.md is a derived index of bank/experience_bank.md, read by the
draft stage to turn its JD-coverage pass into a lookup instead of a per-run re-derivation.
It is never a source of truth: every resume claim traces to the bank, and on any
disagreement the map is stale and the bank wins. Regenerate it whole: read the entire
bank fresh and rewrite the file; do not patch the old map.
Structure (three sections, in this order)
- Requirement lookup: what JDs ask → strongest bank evidence, grouped under theme
headings (platform/backend, full-stack/frontend, data engineering, cloud/DevOps,
databases/performance, ML/AI, systems/embedded, collaboration/communication; adjust
themes to what the bank actually contains). Each line: the requirement as JDs phrase
it → the evidence, tagged with
[role_id], plus the headline facts and a production/ deployed marker where true. - Angle inventory: one entry per bank unit (each contribution and project). Header
carries
[role_id], ownership (sole/partial), and status (production, POC, shipped, competition). Body lists the distinct honest facings the unit supports (e.g. a config platform may face schema design, API design, distributed systems, and IaC), each angle with the exact facts (numbers, tools, tradeoffs, outcomes) behind it. - Absences and standing adjacency candidates: bank-wide absence facts (languages, clouds, platforms, role types the bank does NOT evidence), verified by scanning the whole bank, so the drafter can settle gap verdicts without a full re-scan; then the nearest-honest-evidence candidates for requirements that recur in postings, marked explicitly as per-JD decisions.
Rules
- Facts, never prose. Telegraphic fragments only; no sentence that could be lifted onto a resume. If an entry reads like a finished bullet, break it back into facts.
- Numbers verbatim from the bank. Metrics, costs, scale, dates: copy exactly; never round further or extrapolate.
- Every entry keyed.
[role_id]on everything, so the drafter maps evidence to resume structure without re-reading headers. - Angles must be honest facings, not aspirations. An angle exists only if the bank's facts under it would survive the evaluator's grounding check on their own.
- Absence claims are claims. State an absence only after checking the entire bank, including the projects and skills sections.
- Never launder placeholders. Bracketed TODOs or unfilled fields in the bank are not facts; where a unit has them, note the boundary ("do not claim beyond the above").
- Stamp it. Open the file with the generation date, the regeneration command, and the derived-not-source contract.
When done, confirm the map's unit count matches the bank's (every contribution and project has exactly one angle-inventory entry) and stop.
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.

