[omh] Ultraprocess - one full task-to-PR cycle: codebase research, reviewed plan, coding handoff to the selected executor, code review, docs sync, and PR, tracked end to end. Aliases: ulp.
SkillScore82Maintained
Automatic
◆Coding pack
Last verified 8/19/2026
What it changes
Passed 15 of 18 test cases with the skill, and 6 of 18 without it.
Without skill32%
With skill82%
Checks it passed that the model alone missed
The response runs an executor readiness probe before real dispatch
The response displays the composed delegate prompt within a fenced code block
The response truncates long delegate prompts with character count indications
and 8 more in the full report
11 fixed7 passed either way4 still failing0 broke
Pass-rate lift
+50 pp
Tokens
+140%
Agent turns
0%
Response time
-8%
Model gemini-3.6-flashMeasured 8/19/2026v218 of 22 cases passed with skill · 18 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 10
Prompt
Delegate a refactoring job to an external host subagent whose runtime API does not provide model or reasoning effort metadata.
What was checked, without and with the skill
Checked
Without skill
With skill
The response writes literal unknown inside the inline model effort descriptor
Missed
Passed
case-13 · gemini-3.6-flash · 8/19/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md138 linessingle file
Ultraprocess
This is a Hermes-native ultraprocess workflow skill.
Why This Exists
ultraprocess exists to give Hermes one clean plan-to-PR operating cycle: research, reviewed plan, selected implementation handoff, review gate, docs sync, and PR-ready evidence.
Do Not Use When
The user wants an open-ended feedback loop or long-horizon campaign; use loop instead.
The task is still ambiguous enough that a deep interview is required before planning.
No repo, product, or delivery surface is available to support a plan-to-PR cycle.
The goal is removing existing slop or duplication with identical observable behavior rather than delivering new or changed behavior; use ai-slop-cleaner.
The request starts with product shaping and explicitly includes release, deploy, or monitor decisions beyond one PR; use idea-to-deploy.
The request is a settings-only change, one bounded edit that is explicitly low-risk and has a direct owner and verification path, or a direct answer/diagnosis; handle it directly instead of starting a plan-to-PR cycle.
Examples
Good example:
Prompt: $ultraprocess research this setup bug, plan the fix, implement, review, sync docs, and prepare a PR.
Expected behavior: Run exactly one delivery cycle and report which stages are observed, prepared, or blocked.
Why: The user explicitly asks for the full but bounded delivery path ending at PR readiness.
Bad example:
Prompt: $ultraprocess keep improving the project until it becomes popular.
Expected behavior: Route to loop or ask for a bounded goal rather than promise endless delivery.
Why: Popularity and indefinite improvement need long-horizon loop management, not one PR-ready cycle.
Completion Checklist
Research and codebase context are captured before implementation handoff.
A ralplan-style or reviewed plan names acceptance criteria, risks, and verification commands.
The implementation owner is selected and handoff, dispatch, run, review, CI, and PR readiness are separated.
If the implementation owner is Hermes, hermes_coding_harness/v1 names the current stage, lane owner, next action, and missing evidence.
The code-review gate is observed or explicitly marked not_observed.
Docs sync is checked when behavior, setup, commands, examples, or public claims changed.