capscore-agent / examples /sample_result.md
mathurinacheisoft's picture
Deploy CAPScore (clean recreate)
2a7f679 verified
|
Raw
History Blame Contribute Delete
3.93 kB

CAPScore Report β€” Job capscore_20260612_100000_abc12345

Overall Score: 84/100

Audit type: audit_repository Β· Completed: 2026-06-12T10:04:30Z Β· Duration: 4m 28s


Judging-Aligned Scorecard

Dimension Weight Score Notes
Technical Execution 30% 85.7/100 12/14 checks passed. Repository reproducible; CI pipeline present.
A2A Composability 25% 85.7/100 6/7 A2A checks passed. Callable by agents, typed I/O, pricing defined.
Innovation 20% 80.0/100 4/5 innovation checks passed. Unique use case; 4 claims backed by evidence.
Adoption Readiness 15% 83.3/100 5/6 checks passed. Quickstart present; SLA defined.
Presentation Readiness 10% 80.0/100 4/5 checks passed. No demo video URL provided.

Weighted overall: 0.30Γ—85.7 + 0.25Γ—85.7 + 0.20Γ—80.0 + 0.15Γ—83.3 + 0.10Γ—80.0 = 84.0


Critical Issues

  1. No Agent Store listing URL provided β€” required for hackathon submission
  2. Failure/timeout behavior not documented in CAP integration
  3. LICENSE file not detected in repository root

Top Fixes

  1. Register agent on agent.croo.network and add listing URL to submission
  2. Add failure_modes section to cap-provider/src/schemas.ts
  3. Add MIT LICENSE file to repository root
  4. Add demo video URL (YouTube or Loom) to README and submission
  5. Document SLA timeouts in README and CAP schema

Claim Verification

Claim Status Confidence Notes
This agent provides a working CAP provider. βœ“ supported 92% cap-provider/src/index.ts registers capabilities via CROO SDK
The agent can be called by other agents via CROO SDK. βœ“ supported 88% registerCapabilities.ts exports 3 typed capabilities
The README setup works in less than 10 minutes. ~ weak 65% First-time Docker pull may exceed 10 min on slow connections
The agent delivers verifiable proof packs with SHA-256 hashes. βœ“ supported 95% proof_pack.py builds deterministic ZIP with SHA-256 manifest

Suggested rewrite for weak claim: "The README setup typically works in under 15 minutes with Docker installed."


A2A Calls Made During Audit

Agent Task Status Duration
source-verifier-mock verify_github_source cleared 7s
security-scanner-mock scan_secrets cleared 90s
readme-rewriter-mock suggest_readme_improvements cleared 60s

Proof

  • Result hash: sha256:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
  • Proof pack: proof-pack-capscore_20260612_100000_abc12345.zip
  • A2A calls: 3
  • Generated: 2026-06-12T10:04:30Z

Proof Pack Contents

proof-pack-capscore_20260612_100000_abc12345.zip
β”œβ”€β”€ manifest.json          # Job metadata + file inventory + A2A call log
β”œβ”€β”€ result.json            # Full scorecard in canonical JSON
β”œβ”€β”€ result.md              # Human-readable report (this file)
β”œβ”€β”€ result_hash.sha256     # sha256:<hex> of canonical result.json
β”œβ”€β”€ execution_log.jsonl    # Timestamped worker events
β”œβ”€β”€ attestation.json       # CAPScore Agent statement + log hash
└── evidence/
    └── sources.json       # Repo file inventory + security findings

Verify Locally

# Download proof pack
curl -O http://localhost:8000/jobs/capscore_20260612_100000_abc12345/proof-pack.zip
unzip proof-pack-capscore_20260612_100000_abc12345.zip

# Verify hash
EXPECTED=$(cat result_hash.sha256)
COMPUTED="sha256:$(python3 -c "
import json, hashlib
data = open('result.json').read()
obj = json.loads(data)
canonical = json.dumps(obj, sort_keys=True, ensure_ascii=False, separators=(',',':'))
print(hashlib.sha256(canonical.encode()).hexdigest())
")"
[ "$EXPECTED" = "$COMPUTED" ] && echo "VERIFIED" || echo "MISMATCH"

Generated by CAPScore Agent v0.1.0 β€” CROO Hackathon 2026