agentleFS
IyedNasraIyedNasra/skeptic/SKILL.md

skeptic

Skill1 starsChanged 3 months agoMITView on GitHub

Use when debugging failures, writing tests, starting multi-step tasks, modifying build systems, integrating APIs, or validating assumptions before heavy execution.

6 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions6 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 6

    1. codexstar69/bug-hunter★ 546same name
    2. opentensor/subtensor★ 402same name
    3. sergeeey/Claude-cod-top-2026★ 9same name
    4. maccydee/claude-planning-skills★ 3same name
    5. HumanBean17/agent-mindset-skills★ 1same name
    6. tokyubevoxelverse/skeptic★ 1same 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

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Rule: Treat user hypotheses as unverified inputs.

    line 33
  2. Very specific:

    When completing a high-blast-radius gate, state proof in one line: [Skeptic Gate: observed failure repro -> applied patch -> observed green …

    line 61

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “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

Skeptic

Calibrated skepticism for AI coding agents. Catch unverified assumptions, detect runtime mirages, and stop premise compounding before wasting tokens.

The Seven Traps

1. The Ghost Target Trap (Runtime Mirage)

  • Failure: You edit src/index.ts. The runtime executes dist/index.js, a cached Docker layer, or a shadowed module. You spend turns debugging unchanged code.
  • Rule: Prove code execution. Check build timestamps, inspect watch processes, or inject an observable scalar probe into output.

2. The Tautological Test Trap (The Echo Chamber)

  • Failure: You write the test after writing the fix. The test asserts what the code produces rather than what the specification requires, or mocks out the broken code path.
  • Rule: A test proves nothing until you see it fail. Reproduce the failure on pre-fix code before asserting a fix.

3. The Silent Success Trap (Code 0 Delusion)

  • Failure: A script, migration, or API returns code 0 or status 200. You declare success. The script swallowed an error, matched 0 rows, or returned an empty array [].
  • Rule: Exit code 0 does not prove work. Inspect scalar outcomes: affected rows, non-zero file size, or modified lines.

4. The Premise Compounding Trap (Unverified Foundations)

  • Failure: You assume two databases contain distinct data, an API returns snake_case, or an ID column contains unique keys. You write 300 lines of transformation logic on this premise. The premise fails; you discard the code.
  • Rule: Gate step 0. Run a one-line scalar check on core premises before writing pipeline logic.

5. The Sycophantic Diagnosis Trap (Echoing User Guesses)

  • Failure: The user suspects a bug in the cache layer. You rewrite the cache without checking logs. The real bug is an unhandled rejection in the transport client.
  • Rule: Treat user hypotheses as unverified inputs. Check stack traces and telemetry before changing code.

6. The Phantom Symbol Trap (Hallucinated Contracts)

  • Failure: You assume a library exports a convenience method (fs.existsAsync, stripe.customers.bulkUpdate). You build around an API missing from the installed package version.
  • Rule: Inspect type definitions or local exports before building on third-party API assumptions.

7. The No-Repro Trap

  • Failure: You edit code and declare a fix based on deductive reasoning without running a reproduction script.
  • Rule: Without reproduction, you have no verified fix. Observe the failure state before changing code.

The Asymmetry Rule

Match verification effort to blast radius:

  • Low blast radius (routine UI styling, local comments, helper functions): Proceed. Run no verification steps.
  • High blast radius (data migrations, multi-source pipelines, auth and build logic): Run an explicit falsification check.

Rationalization Table

ExcuseRealityAction
"The script exited with code 0."Code 0 masks empty runs and caught exceptions.Check scalar outputs (count > 0).
"The test passed, so the fix works."Tests written after fixes often test tautologies.Run test against pre-fix code; confirm it fails.
"The user identified this root cause."Users report symptoms; logs show facts.Read logs and traces before editing.
"The file paths differ, so the data differs."Different paths often link to duplicate fixtures.Compare checksums (cmp -s).
"This library method must exist."Models blend API versions.Check type declarations or grep package exports.

Gate Receipts

When completing a high-blast-radius gate, state proof in one line: [Skeptic Gate: observed failure repro -> applied patch -> observed green state]

References

  • references/cognitive-traps.md: Failure mode analysis and case studies.
  • references/falsification.md: Reproduction and falsification patterns.
  • references/scalar-probes.md: One-line shell and SQL recipes (<20 tokens).
  • references/tripwires.md: Passive anomaly signatures and root causes.

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.

Reviews can't be read right now.

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.