Install
$ agentstack add skill-jasontang-ai-acp-omission-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
What this skill is for
Use this skill to catch missing issue coverage before a routed digest or briefing is released.
Constitutional purpose
Protect material issue coverage and minority-salience against compression loss, loud redundancy, and superficial completeness claims.
Failure modes targeted
- Material issue omission
- Stakeholder omission
- Late-arriving or minority-salience signal burial
- False completeness claims
When to use it
- After routing but before digest release.
- When Engine V2 reports missing issue clusters.
- When a minority or late-arriving signal may have been buried.
When not to use it
- Do not use it to add irrelevant material for balance theater.
- Do not use it as a generic summarizer.
- Do not redefine canonical ACP criteria.
Inputs expected
- Cycle JSON or issue map
- Contribution records
- Routing decisions or digest items
- Current omission critic output when present
Required evidence for valid use
- Selected items and excluded-but-relevant material
- Enough issue clustering evidence to compare coverage rather than guess
- A concrete output under review, not only a vague concern about representativeness
Preflight / prerequisites
- Confirm the output under review is an intervention or briefing artifact.
- Identify the selected items and excluded-but-relevant items.
- Confirm available issue clusters and unresolved questions.
Workflow
- Compare selected items against issue clusters.
- Identify important missing issues and stakeholders.
- Check whether loud/redundant comments displaced useful dissent.
- Recommend release, revise, or escalate.
Decision rules / judgment criteria
- Missing material issues matter more than perfect topical balance.
- Late minority-salience contributions deserve explicit review.
- Unresolved questions should survive compression when decision-relevant.
Escalation rules
Escalate when a missing issue affects the decision, when a stakeholder class is absent from selected items, or when the digest implies completeness without coverage.
Available scripts
scripts/smoke.mjs: emits the skill identity and expected review surfaces for audit and composition checks.
Produced artifacts
- Missing issue list
- Underrepresented stakeholder notes
- Revise/release/escalate recommendation
Allowed claims
- That specific issues or stakeholder signals appear missing relative to the supplied evidence
- That release should pause because selected material overstates completeness
- That redundancy displaced relevant dissent or unresolved questions
Forbidden claims
- That every omitted item must be included for fairness theater
- That omission review alone solves representativeness
- That a missingness suspicion counts as proof without supporting evidence
Human override points
- Materiality judgment on whether an omission changes the decision
- Manual selection repair before release
- Facilitator veto when completeness claims exceed coverage evidence
Outputs
- Missing issue list
- Underrepresented stakeholder notes
- Revise/release/escalate recommendation
Failure handling
If issue-map evidence is missing, mark the review incomplete rather than inventing clusters.
Trust / safety notes
Omission critique can increase complexity. Use it to protect material coverage, not to force every voice into every digest.
Composition notes
Use with epistemic-routing, digest-and-explanation, and abstention-escalation.
Examples to inspect next
Read examples/casebook.md for omission review cases.
Evaluation hooks
Run scripts/smoke.mjs and npm run skills:audit.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: jasontang-ai
- Source: jasontang-ai/acp
- License: Apache-2.0
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.