// HR & Recruiting

Evidence-Bound Resume Builder

Name: Evidence-Bound Resume Builder
Brand: JustHandled Labs
Price: 12 USD
Availability: InStock

Prepare a role-specific resume and matching cover-letter draft from your permitted career facts, with excluded claims and shared provenance.

Get it on Agensi — $12 Find another skill $12 one-time

Prefer Gumroad? Get Evidence-Bound Resume Builder for $12 USD on Gumroad. Review checkout for any applicable tax or currency conversion.

Format: Downloadable agent skill, not a browser app. You need an AI agent that can load its instructions and use the required tools.

Requirements: Python 3.10+ and the complete downloaded package. The helper uses the standard library; Windows Python 3.12.14 tested. Agent-host installation is not independently tested.

You get: Resume and cover-letter drafts, source provenance, requirement coverage and questions for your review.

Licence and purchase details

JustHandled Labs paid skill license; included LICENSE.txt governs

Review the offer and download options at the linked marketplace. Buying a skill does not include an AI-agent subscription or setup service.

Your first useful result

Start with the included fictional example before sharing any career information.

Show the first-run walkthrough

Prepare. Extract the whole package, keeping scripts, references and fixtures together. You need Python 3.10 or newer. Open a terminal in that package folder; the example below creates a new output folder there.

Run from the package folder.

python scripts/run.py fixtures/practice.json --output first-resume-practice

Review the output. Expect REVIEW, not a failure: two supported statements appear in resume-draft.txt and cover-letter-draft.txt. The invented “Increased revenue by 73%” claim is excluded, and missing budget-leadership evidence is flagged. Read candidate-questions.md and the provenance files before changing the example.

The helper assembles approved source statements; it does not independently verify your career or produce a finished page design. Your agent may process prompts remotely. Review its privacy settings before supplying personal information. Use a new output folder for each run; its parent folder must already exist.

Included synthetic example re-run on Windows with Python 3.12.14. This verifies the local helper, not automatic installation in every agent host.

Prefer to ask your agent? Copy this starting prompt

Read this package’s SKILL.md and references/input-contract.md. Help me prepare a draft for my target role using only sources I explicitly permit. Preserve volunteer labels, caveats and negation. Ask me to approve the source units, create the required input packet, then run the helper into a new output folder. Return both drafts and the unresolved questions. Do not invent facts, contact anyone or submit anything.

Make your experience clear. Keep it yours.

28 seconds · illustrative animation · mechanical sound effects, no narration. On a small screen, use fullscreen or read the example below.

Read the film transcript

A supplied reporting-assistant role asks for Python reporting and budget leadership. Volunteer Python reporting experience supports the first requirement. A proposed draft adds “Increased revenue by 73%,” but that claim is absent from the permitted source. The helper excludes it. Before and after views retain the volunteer statement with its source locator and remove the unsupported metric. Budget leadership becomes a question for the candidate. A cover-letter draft is an additional output. Review both drafts; supplied records are not independent career verification.

A readable version of the example

Permitted source:

I built Python reporting tools for our weekly volunteer rota.

Proposed addition, excluded: Increased revenue by 73%.

Retained: the supplied volunteer experience, linked to notes:line-2. No revenue figure is added.

Still unresolved: evidence for budget leadership. The helper returns REVIEW, produces two supported statements in each draft, excludes c2 and flags r2. The film illustrates one retained statement; the complete outputs include both. It does not design the finished resume page.

Review the current offer on Agensi

Use it to

For an applicant tailoring to a supplied role, prepare resume and cover-letter drafts from one permitted fact set so unsupported claims are excluded from both, with a shared provenance file.
Prepare your next application or practice session using your own approved facts.
Return to a saved packet without losing earlier work.

What it returns

Two editable drafts, plus the questions, coverage and source trail needed to review them.

See every output file

application-provenance.csv
application-review.md
candidate-questions.md
claim-provenance.csv
cover-letter-draft.md
cover-letter-draft.txt
findings.csv
manifest.json
original-resume.txt
report.md
requirement-coverage.csv
resume-draft.md
resume-draft.txt
summary.json

A representative input and result

Supplied volunteer reporting experience reaches the cover letter. An unsupported revenue claim does not.

Inspect the complete fictional input and draft

fixture-backed sample

input {
  "original_resume": "I built Python reporting tools for our weekly volunteer rota.\nI checked the rota against availability notes and documented the corrections for the coordinator.",
  "job_description": "Python reporting role",
  "sources": [
    {
      "id": "s1",
      "text": "I built Python reporting tools for our weekly volunteer rota.",
      "locator": "notes:line-2",
      "permitted": true
    },
    {
      "id": "s2",
      "text": "I checked the rota against availability notes and documented the corrections for the coordinator.",
      "locator": "volunteer-notes:4",
      "permitted": true
    }
  ],
  "claims": [
    {
      "id": "c1",
      "text": "I built Python reporting tools for our weekly volunteer rota.",
      "source_id": "s1",
      "requirement_ids": [
        "r1"
      ]
    },
    {
      "id": "c2",
      "text": "Increased revenue by 73%.",
      "source_id": "s1",
      "requirement_ids": [
        "r1"
      ]
    },
    {
      "id": "c3",
      "text": "I checked the rota against availability notes and documented the corrections for the coordinator.",
      "source_id": "s2",
      "requirement_ids": [
        "r1"
      ]
    }
  ],
  "requirements": [
    {
      "id": "r1",
      "text": "Python"
    },
    {
      "id": "r2",
      "text": "Budget leadership"
    }
  ],
  "application": {
    "role": "Reporting assistant",
    "company": "Example Community Centre",
    "job_locator": "supplied-job:1",
    "claim_ids": [
      "c1",
      "c2",
      "c3"
    ]
  }
}

result Dear hiring team,

I am applying for the Reporting assistant role at Example Community Centre.

I built Python reporting tools for our weekly volunteer rota.

I checked the rota against availability notes and documented the corrections for the coordinator.

Thank you for considering my application.

Access and approval boundaries

Read one local input JSON file.
Write a new local output directory; no network or external actions.

Known limitations

Extractive source-unit assembly; candidate and agent improve presentation without inventing facts.
No automatic career verification, ATS score, PDF resume designer or application submission.
Role/company and fact relevance are supplied, not independently authenticated.

Questions

What do I receive?

A downloadable skill with a local Python helper, input contract, synthetic fixtures, tests and sample reports.

What do I need to run it?

Python 3.10 or newer and your authorized normalized JSON records. The helper uses the standard library; Windows CLI execution was tested.

Does it connect to my HR software?

No. You supply a local export normalized to the documented schema. There is no ATS or HRIS integration.

Will it apply, contact employers or judge my suitability?

No. It helps you prepare your own materials or practice. It never sends or submits, infers sensitive traits or predicts job outcomes.

What does READY mean?

No encoded finding in the supplied scope. It does not prove source truth, comprehensive readiness, legal compliance or approval to act.

What are the main limits?

Extractive source-unit assembly; candidate and agent improve presentation without inventing facts.

Where does my data go?

The Python helper stays local. Your agent host may separately process prompts; use synthetic data unless the owner authorizes sensitive data processing.