agentsclimarketplace
Sign in

Piensalo verify

Skill ralfyishere/piensalo/skills/piensalo-verify

Independent verification of a candidate answer against domain verifier criteria: deterministic checks first, adversarial probes, evidence-level grading, disqualifier sweep, per-criterion verdict table. Use when asked to check, grade, review, or attack a finished answer or artifact - yours or another agent's - before it is trusted or shipped. Trigger phrases: 'verify this', 'is this answer right', 'review before we ship', 'grade this output'.From its SKILL.md

Install
npx -y skills add ralfyishere/piensalo --skill piensalo-verify

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 27 days oldThe repository was created 27 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its file declares

Copied from the file, not written here

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

2.6 KB, 452 tokens by cl100k_base, as published. Nobody here has run it

piensalo-verify

The verification workflow of Piénsalo. Given a task and a candidate answer, grade the candidate - do not fix it. Verification is adversarial: your job is to find where it fails.

Primary workflow

  1. Pick the domain verifier - software, mathematics, research, strategy, invention, writing, or generic if none fits. Criteria for every domain are in references/verification.md.
  2. Disqualifier sweep first - check the domain's disqualifiers; any hit ends the review with verdict FAIL and the quoted evidence.
  3. Run deterministic criteria - execute or mechanically check everything checkable (run tests, recompute numbers, diff scope). Quote actual output; a prediction is not a result.
  4. Run adversarial criteria - construct probes the author visibly did not try (empty/boundary/hostile inputs, counterexamples, contradicting sources); predict the outcome before executing each probe.
  5. Grade evidence criteria - for judgment-level criteria, classify the candidate's support as verified fact, inference, assumption, or guess; unsupported load-bearing claims fail the criterion.
  6. Verdict table - one row per criterion: criterion | kind | evidence | verdict. Overall verdict fails if any disqualifying criterion fails. State what could not be checked here and why.

Conditional moves

  • When a deterministic criterion needs tools you lack, mark it NOT-RUNNABLE-HERE - never infer a pass.
  • When the candidate spans domains, apply generic criteria plus each touched domain's disqualifiers.

Output contract

Per-criterion verdict table, disqualifier findings quoted verbatim, list of unchecked criteria with reasons, and an overall PASS/FAIL/NOT-FULLY-CHECKABLE verdict.

Delivery notes (small-model packets)

If delegating steps to a smaller model: one bounded objective per packet, all inputs named explicitly, facts separated from instructions, uncertainty marked 'UNCERTAIN: ...' rather than invented away.

Before answering

Check the draft against references/failure-checks.md (the failure-mode catalog) and references/verification.md (domain verifier criteria). Repair procedures live in references/contracts.md.

What ships with it: 9 files

53.8 KB alongside SKILL.md

evals/

examples/

references/

Keep looking

Skills are one crate of 326,851. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.