Install
$ agentstack add skill-akshay-diwadkar-skills-critic ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Critic
Contract
Require a concrete proposal or result. Challenge one load-bearing claim whose failure changes the owning decision. Do not solve the task, replace its specialist, or manufacture authority.
Challenge the supplied target rather than reopening goal selection or routine domain work. Return decision-readiness problems to the owning task; Think may handle them when available but is never required for Critic to operate.
Frame and independence
Bound input to target, goal/constraints, authoritative evidence, material assumptions/unknowns, consequence, and stop condition. Prefer proposal plus evidence over the proposing agent's transcript. Never request or persist chain-of-thought.
Prefer a fresh bounded worker or context when proportionate. Otherwise perform one adversarial pass without claiming independent multi-agent review or simulating a persona panel.
Runtime is temporary, read-only, and artifact-free. Owners set triggers/review breadth, re-verify evidence, decide, and record only accepted consequences in authority; Critic creates no state. Missing user authority uses Elicit/direct question, not a Critic disposition. Router/Shepherd never route or persist Critic.
Challenge
Seek the strongest decision-changing contradiction, assumption, counterexample, alternative cause, omission, false proof, consequence, scope mismatch, authority defect, stale premise, or outcome gap. Reject style/preference criticism.
When the lens or discriminator is unclear, read [challenge protocol](references/challenge-protocol.md). Stop at one material objection, no material objection, or irreducible evidence gap. Do not invoke critic again on unchanged target and evidence.
Result
Always return Target, Disposition, and Load-bearing claim.
material-objection: addStrongest challenge,Evidence / counterexample,Why it matters, andSmallest discriminator or repair.insufficient-evidence: add the unresolved material possibility, consequence, and cheapest discriminator; state when none exists.no-material-objection: state only that the bounded pass found no decision-changing defect. Never present this as proof of correctness.
Return only decision-useful information. Create no critique artifact, confidence score, vote, transcript, or state. Return control to the owning task.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: akshay-diwadkar
- Source: akshay-diwadkar/skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.