Showcase mode. Everything here is pre-generated evidence produced by Airlock before this site was built. Nothing runs here, nothing is uploaded, and no verdict can be created or changed on this site. The real product is the local operator platform.
Big Red — the case study
Genuine Airlock output over the repository's approved illustrative Big Red case-study fixture, not an assessment of the current full Big Red repository.
PASS_WITH_CONDITIONS — Proceed, with conditions recorded
Recorded by Airlock on 2026-08-19T05:06:27.506Z; displayed, not computed, here.
Airlock recommendations do not constitute human deployment approval.
No blocking findings. 3 condition(s) must be satisfied as part of the deployment itself, and confirmed before the agent is treated as live.
The conditions Airlock recorded
▲ Declared egress allowlist is not used as an allowlist anywhere in the scanned source
Finding
Severity
Condition (DEPLOYMENT_CONDITION)
Id
egress-uncorroborated
Check
AL-018
Source
deterministic
The manifest declares default-deny egress limited to api.anthropic.com, api.tavily.com, api.firecrawl.dev, www.staples.com, www.officedepot.com. No scanned file uses any of those hosts in a decision or an allowlist collection. The restriction may be enforced elsewhere — a proxy, a gateway, infrastructure outside scanPaths — but nothing in the declared source supports it. Absence is weaker evidence than contradiction, so this is a condition, not a blocker.
Remediation, as the engine recorded it
Point scanPaths at the code that performs outbound calls, or record where egress is enforced if it is enforced outside this tree.
Evidence lines, verbatim
searched 3 file(s) under ../samples/big-red
allowlisted hosts used as an allowlist: 0
▲ Adversarial review was skipped by the operator
Finding
Severity
Condition (DEPLOYMENT_CONDITION)
Id
review-skipped
Check
AL-000
Source
deterministic
This gate ran deterministic checks only. Injection, exfiltration, excessive-agency and confused-deputy risks were not assessed.
Remediation, as the engine recorded it
Re-run without --no-review before treating this verdict as a complete assessment.
Evidence lines, verbatim
reviewMode=skipped
effectiveTier=moderate
▲ Rollback path has not been tested
Finding
Severity
Condition (DEPLOYMENT_CONDITION)
Id
rollback-untested
Check
AL-009
Source
deterministic
An untested rollback is a plan, not a control. The first time you run it should not be during an incident.
Remediation, as the engine recorded it
Execute the rollback once in a non-production environment and record the result.
Evidence lines, verbatim
No evidence lines recorded.
Clearing every item above does not guarantee a pass on a later run, and does not make a deployment safe.
What corroboration means here
A passing corroboration check means not contradicted. It never means verified.
Kill switch
corroboration: corroborated
"PRICE_LOOKUP_ENABLED" found in: samples/big-red/.env.example, samples/big-red/lib/kill-switch.ts
files searched: 3
scan issues: 0
Provider ceilings
files searched: 3
scan issues: 0
tavily: corroborated in samples/big-red/lib/budget.ts
firecrawl: corroborated in samples/big-red/lib/budget.ts
anthropic: corroborated in samples/big-red/lib/budget.ts
Secret scan
paths declared: ../samples/big-red
files scanned: 3
patterns applied: 12
hits: 0
coverage issues: 0
The two Big Red runs in this corpus used identical inputs. Their checks, findings and verdict are identical; their generation times and digests are not. See the pair →