Passed 20 of 20 test cases with the skill, and 11 of 20 without it.
Without skill55%
With skill100%
Checks it passed that the model alone missed
Restricts writing verdict.v2 exclusively to Validate.
Requires case-insensitive path comparisons when resolving scopes.
Rejects non-empty write_scope.exclude settings in dispatch_once.py.
and 7 more in the full report
9 fixed11 passed either way0 still failing0 broke3 errored
Pass-rate lift
+45 pp
Tokens
+47%
Agent turns
0%
Response time
-51%
Model gemini-3.6-flashMeasured 10/10/2026v620 of 23 cases passed with skill · 20 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 9 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 9
Prompt
After observing candidate checks on `services/order.py`, the Scribe session logged the evidence. Can the Scribe or Implementer write `verdict.v2` directly?
What was checked, without and with the skill
Checked
Without skill
With skill
Restricts writing verdict.v2 exclusively to Validate.
Missed
Passed
case-21 · gemini-3.6-flash · 10/10/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md119 lines7 filesincludes scripts
Agent Native
Launch and observe caller-selected agent sessions as explicit roles, and return
runtime facts per packet. Execution does not validate output, and the runtime
never becomes AgentOps lifecycle authority: an adapter cannot select AgentOps
semantics, issue a binding verdict, or turn factory completion into delivery or
validation proof. Orchestrate decides what to
dispatch and when; this skill launches and observes.
Before launch, and while running
Validate the whole batch first. Every packet needs its selected executor,
packet identity, all transitive effects and a canonical workspace-relative
write scope in separate isolation. If any packet is invalid, launch none:
starting the valid ones and fixing the rest later is a partial launch.
Compare canonical scopes. Resolve symlinks, normalize paths and compare
case-insensitively so an alias cannot hide a collision. A lexical check alone
cannot prove symlink or runtime isolation.
Delivery is not engagement. A delivered prompt or an acknowledgement
proves transport only; reading it as a working worker is prompt-send
optimism. Prove engagement from observable state: session output, tool
activity, changed files.
Capture before restart. Before a nudge, replacement or restart, capture
the worker's observable state. A restart destroys the evidence of why it
stalled, and rescue is usually cheaper than rerun. Silence or impatience alone
does not justify a restart; any replacement stays within authority and
remaining bounds.
Dispatch once. Keep each packet's identity with its result. An executor
error is a result, not a retry trigger; repair and follow-up are the caller's
authority.
Roles
Orchestrator: passes focused intent, scope and evidence references, names
the integration/final-review owner, and reports runtime facts. Retrieve extra
history only for a consequential uncertainty; a fresh context is not
necessarily small. A new goal does not clear history or renew spent bounds.
Implementer: may modify only its packet's declared subject.
Validator: receives exact candidate content in a fresh, read-only context.
It may supply judgment to Validate; only Validate writes verdict.v2.