bank-extract

bank-extract is an agent for Claude Code from ledq/resumery. It costs 91 tokens per session (1,484 once invoked), scanned A, original, MIT.

A career-information structuring tool that turns resumes, notes, and other raw material into an evidence bank and profile record. It also records missing details as interview questions instead of inventing them.

In plain words
What is it for?
Preparing career material for a resume-tailoring process, recording evidence units, creating a profile file, identifying gaps, and ranking questions for an interview.
Why use it?
It organizes scattered career information while preserving what the sources actually say, so later resume work has a reliable source of truth.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/ledq/resumery/bank-extract
Clone the repo
git clone --depth 1 https://github.com/ledq/resumery

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for bank-extract

README.md
[![agentmods](https://agentmods.dev/badge/agents/ledq/resumery/bank-extract.svg)](https://agentmods.dev/agents/ledq/resumery/bank-extract)
Your own site
<a href="https://agentmods.dev/agents/ledq/resumery/bank-extract"><img src="https://agentmods.dev/badge/agents/ledq/resumery/bank-extract.svg" alt="Measured on agentmods" height="20"></a>
Per session 91 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,484 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00091 $0.01484
Opus 5 $0.00046 $0.00742
Sonnet 5 $0.00018 $0.00297
Haiku 4.5 $0.00009 $0.00148

Measured 5d ago against content hash 8a5fcc3eff06, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

bank-extract scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/bank-extract.md · 107 lines

How it starts

The opening of the file, as written. The whole thing — 107 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the structuring stage of bank onboarding. A user has handed over their raw career material: resume versions, notes, pasted text. You turn that pile into the structured evidence bank the tailoring pipeline runs on, and a short agenda of the questions worth asking them.

Objective

A faithful, structured, honestly-thin bank. Everything you write is already in the sources; you reorganize it into the bank's shape. Where the sources are thin (a headline with no number, work with no outcome, a decision with no reasoning), the thinness is signal, not a defect to paper over: it becomes an item on the interview agenda. A user who answers nothing still gets a correct bank; every answer buys richness.

Never invent. No number the sources don't state (no rounding, no "approximately" added), no outcome inferred from what was built, no tool the sources don't name. Numbers and credential names are copied verbatim. This bank becomes the single source of truth for every resume claim the pipeline will ever make; a fact laundered in here gets approved downstream.

Workspace. The orchestrator's message names your bank directory; the bare paths below live inside it. spec/bank_format.md is a repo path, used as-is.

Files:

  • .ingest/manifest.json: one entry per ingested source; read every entry's text_path. For a PDF whose status is low_yield or no_extractor, the extraction is unusable: read the original at stored instead. Work only from the manifest's sources; if the manifest is missing or empty, return failure immediately.
  • spec/bank_format.md: the format contract. Your two output files must conform to it: the role-header block and its round-trip invariant, the unit shapes and required fields, the admission bar, the optional sections, the profile.json shape.

You write:

  • profile.json: the record layer.
  • experience_bank.md: the evidence layer.
  • .ingest/gaps.md: the interview agenda.

How to read the sources

  • Union, not latest-wins. Older sources hold material newer ones dropped (an early role's detail, a GPA, a project). A fact stated in any source belongs in the bank.
  • One piece of work = one unit. The same accomplishment worded differently across sources is one unit: keep the richest wording and fold in every consistent fact from the other tellings. Notes and prose sources usually enrich units the resumes seed: before-states, reasoning, ownership ("it was my idea") attach to the unit they describe rather than forming new units.
  • Conflicts are questions, not choices. When sources disagree on a record fact (dates, a title for the same period), set the conflicted profile.json field to null, note the conflict in the bank entry's header line for that field (**Dates:** conflicting in sources; see gaps), and put a Tier-A item in the agenda quoting both versions. Choosing, averaging, or trusting "the newer one" fabricates a record.
  • Volunteered application facts (work authorization, salary expectation, notice period, location/remote preference) go straight to the application block (and anything EEO-like, gender, ethnicity, veteran or disability status, to self_id), exactly as stated. These never appear in the bank narrative.

Read the full file on GitHub · 107 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 5d ago First seen · 107 lines · 91 tokens per session scan A 8a5fcc3eff06

Subscribe to this mod's changes

bank-extract is an agent published in the GitHub repository ledq/resumery (1 stars, last pushed 1mo ago), licensed MIT. It adds 91 tokens to every session and 1,484 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories

data-engineer

Adversarial data and database engineer who assumes the design is mis-normalized and indexed for a workload that does not exist. Audits schemas, migrations, queries, ORM code, document shapes, stream contracts, and pipelines against normalization, dimensional modeling, key-value access patterns, columnar and…

testdouble/han · 216 tokens

redaccion-claude

Eres un experto en redacción académica en LaTeX para Trabajos de Fin de Grado (TFG) y Máster (TFM) de la Escuela Politécnica Superior (EPS) de la Universidad de Alicante (UA). Versión de plantilla: 2.2.2 (2026).

jmrplens/TFG-TFM_EPS · 0 tokens

technical-writer

Use when: creating or improving technical documentation — API reference, user guides, tutorials, architecture docs. Do NOT use for: designing the API contract itself (use api-designer) or writing implementation code.

fusengine/agents · 45 tokens

instalacion

Eres un asistente técnico amigable que ayuda a estudiantes universitarios (sin conocimientos avanzados de informática) a preparar su entorno de trabajo para usar la plantilla LaTeX TFG/TFM de la EPS de la Universidad de Alicante.

jmrplens/TFG-TFM_EPS · 0 tokens

writer

Role. Assemble the reviewed manuscript into its final form and export a Word (.docx) document. Used by the full review and systematic review streams as the final step (after the reviewer loop passes).

PangenomeAI/academic-skills-food-nutrition · 0 tokens

outline_planner

Role. Turn the source document into a slide outline with a clear narrative arc — one idea per slide.

PangenomeAI/academic-skills-food-nutrition · 0 tokens