agentleFS
calebnewtonusccalebnewtonusc/apply-kit/.claude/skills/honesty-guard/SKILL.md

honesty-guard

Skill0 starsChanged 36 days agoMITView on GitHub

Never fabricate a fact about the applicant, keep numbers consistent across every document, and respect org AI bans. Use during drafting, /truth, and before any submission.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When a draft needs a detail you do not have:

    line 16
  2. Very specific:

    Run this check as part of /truth:

    line 36
  3. Distinctive:

    Ask which one is defensible, and make sure the answer includes what it counts.

    line 42
  4. Very specific:

    Write the chosen version into you/PROFILE.md under Canonical numbers, and use it everywhere without exception.

    line 44
  5. Distinctive:

    Not integrity exactly, but the same class of avoidable loss.

    line 73

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Honesty guard

Two failure modes, both expensive, both preventable.

1. Fabrication

Never invent a number, a title, a date, an outcome, a name, or a feeling. Not as a placeholder you plan to fix. Not because it is probably about right. Not because the sentence reads better with a number in it.

When a draft needs a detail you do not have:

I ran the event for [NEED: how many attendees] students across
[NEED: how many schools, and are you counting schools or orgs?] campuses.

Then list every [NEED:] at the bottom of the draft file. A draft with six markers is a good draft. A plausible invented figure is a failure, and it is the kind that surfaces in an interview when someone asks "how did you get that number?"

The same rule covers feelings and motives. Do not write that something moved them, or that they realized anything, unless they said so. Write [NEED: what did you actually think when that happened].

2. Numbers that drift

The same claim appears on the resume, in four essays, and in the interview. If it grows between documents, that is what gets caught.

Run this check as part of /truth:

  1. Pull every number, title, and date from you/PROFILE.md, the resume in you/uploads/, and every file in drafts/.
  2. Group by claim.
  3. Report any claim with more than one value, in a table, showing where each came from.
  4. Ask which one is defensible, and make sure the answer includes what it counts. "Twenty orgs" and "eleven schools" can both be true and are not the same claim.
  5. Write the chosen version into you/PROFILE.md under Canonical numbers, and use it everywhere without exception.

A smaller number they can walk someone through beats a bigger one they cannot. This is also self-interest, not just integrity: orgs that score humility are exactly the ones where an inflated figure does the most damage.

Claims to flag even when true

Some true statements read bigger than the underlying thing, and the interview is where that gets tested. Flag these for the applicant to decide on:

  • A plan described in the past tense. "Secured seven clients for spring" when the clients are a target. An interviewer will ask them to walk through an engagement that has not happened.
  • A round accuracy figure on a small dataset. Invites "what was your split?"
  • A multiplier with no baseline. "6.3x more meetings" invites "from what to what?"
  • A tool listed as a skill they would not want to be quizzed on. Some applications ask which technical skill you want to be evaluated on. Anything on the page is an invitation.
  • A title whose scope changed. Founder, president, lead. If someone else runs it now, the tense matters.

Say it once, give the reasoning, and accept their answer. They know their own work better than any file in this repo does. If they say a claim is accurate, it is accurate, and the right move is to correct the repo rather than restate the doubt.

3. Format requirements are part of honesty too

Not integrity exactly, but the same class of avoidable loss. Every org file has a Format section. Before submitting, verify:

  • File type. Almost everyone wants PDF, and a .md or .pages file fails silently.
  • Filename convention. These differ per org and often include commas or underscores in a specific order.
  • Page limit. If an org says one page, it means one page.
  • Link permissions. If they want a link, open it in a private window first. Orgs that warn twice about permissions do so because people keep failing it.
  • Word limits, counted, per answer.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.