verifiable-orchestration — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited verifiable-orchestration (Agent Skill) and scored it 92/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 2 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 2 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
korgex can split work across subagents and record the whole run as one connected, tamper-evident causal DAG — replayable and auditable (korgex trace/verify). Pick the right shape:
Agent calls inONE turn and they fan out concurrently (results return in order; a crashed sibling is isolated). Use for "explore these 3 areas", "review these 4 files".
nodes withdeps (e.g. explore → plan → implement → review): independent nodes run in parallel, a node waits for its deps, and if a node fails its dependents are skipped rather than run against a broken precondition. The whole run — including the failure topology — is one verifiable DAG.
Lock intent first, for non-trivial runs:
Orchestrate — the agreed {goal, constraints, acceptance_criteria}.It's recorded as an immutable, hash-chained spec.seed the whole run anchors under, so korgex why/trace walk any result back to the spec it was meant to satisfy and korgex verify proves the spec wasn't altered after the fact. Use it whenever the "what we agreed to build" should be pinned before work starts.
Rules:
(hard-enforced) — keep the graph at the top level.
read-only; code gets the full set).
root_seq (and seed_seq) — trace from there toaudit exactly what each subagent did and why.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.