<< All versions
Skill v1.0.0
currentLLM-judged scan90/100oaslananka/zaptrace/zaptrace-proof-pack-review
──Details
PublishedOctober 21, 2026 at 08:26 AM
Content Hashsha256:0974dbccc59a94ca...
Git SHAc77daf519274
──Files
Files (1 file, 2.7 KB)
SKILL.md2.7 KBactive
SKILL.md · 78 lines · 2.7 KB
version: "1.0.0" name: zaptrace-proof-pack-review description: Create or review ZapTrace proof-pack evidence, verify artifact identities, and separate automated gate results from human engineering approval.
ZapTrace Proof-Pack Review
Use this skill when the user asks to create, inspect, compare, or explain ZapTrace proof-pack evidence for a local design.
When to use
Use this skill for:
- Listing available proof checks
- Running bounded proof checks for a parsed design
- Reviewing proof artifacts, hashes, assumptions, skips, and gate outcomes
- Comparing proof evidence across revisions
- Preparing a human-review handoff
Do not use a passing proof pack as a fabrication, safety, compliance, or production-readiness guarantee.
Required inputs
Collect:
- Parsed design identity or local design file
- Proof-pack directory or requested output directory
- Required checks and policy context
- Source revision and environment identity when reproducibility matters
- Known assumptions and approved skips
MCP and CLI workflow
- Use
proof_list_checksto discover the available checks. - Parse and inspect the design with
design_parse_fileanddesign_inspectwhen no design identity exists. - Run
proof_run_designfor a design-bound proof result, orproof_runfor the supported proof workflow. - Inspect related lifecycle records with
audit_list_eventswhen decisions or mutations must be traced. - Verify that reported artifacts carry stable hashes or evidence identities and that the source revision is recorded when available.
- Keep failed, skipped, unsupported, degraded, and human-review-required results distinct from passing results.
- For command-line reproduction, use the documented
zaptrace proofcommands from the source checkout and retain the machine-readable report.
Quality checks
The review must verify:
- Design or source revision identity
- Check inventory and policy identity
- Artifact hashes and referenced files
- Assumptions, warnings, skips, and unsupported domains
- Automated gate result
- Human-review or fabrication-review status as a separate field
- Reproduction command or environment notes when requested
Failure modes
Stop and report clearly when:
- The proof pack is incomplete or its referenced artifacts are missing
- Hashes or source identities do not match
- A result is stale relative to the design revision
- A skipped or unsupported check is represented as passing
- The user asks for an approval that the evidence does not support
Output format
Return:
- Evidence identity
- Design/source identity
- Check results
- Artifact integrity summary
- Assumptions and skips
- Automated gate status
- Human-review status
- Reproduction notes
- Non-claims