Content
67%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body provides a well-sequenced, largely executable operating manual for the ultragoal workflow with explicit gating and blocker-recovery logic. Its main weaknesses are redundancy — the shell-vs-/goal limitation and parallel-session caveats are each stated multiple times — and placeholder-heavy examples that stop just short of copy-paste ready.
Suggestions
State the 'shell cannot invoke or mutate Claude /goal state' limitation once (e.g., in <Important_Limitations>) and reference it briefly elsewhere instead of repeating the full explanation in four sections.
Consolidate the parallel-session guidance in <How_To_Use> and '## Parallel session caveats' into a single section, or move the flag/reference detail (workspace anchor resolution order, session id sources) into docs/REFERENCE.md.
Provide one complete, non-placeholder example of --claude-goal-json and --quality-gate-json payloads so the checkpoint and blocker commands are fully copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient and tool-specific, but the 'shell cannot invoke or mutate Claude /goal state' limitation is repeated in <Purpose>, <Do_Not_Use_When>, <How_To_Use> step 2, and <Important_Limitations>, and parallel-session guidance appears twice (inside <How_To_Use> and again under '## Parallel session caveats'). These redundancies could be tightened into one authoritative statement, so it sits between 'mostly efficient' (3) and 'efficient' (4). | 3 / 5 |
Actionability | Commands are concrete and executable ('omc ultragoal create-goals --brief-file plan.md', 'omc ultragoal checkpoint --goal-id G001-... --status complete --evidence ...', 'omc ultragoal record-review-blockers ...'), including flag-level detail and in-session agent instructions. Not 5 because some examples rely on placeholders (G001-..., '{"goal":{"objective":"..."}}', evidence strings) whose full schemas are not shown. | 4 / 5 |
Workflow Clarity | A clear numbered 1–5 sequence covers create, start/resume, checkpoint, blocker handling, and inspection, with an explicit failure path ('If the final review is not clean, do NOT mark complete. Record blockers') and validation rules ('rejected without state mutation', quality-gate evidence 'all clean'). Not 5 because per-step output validation and recovery loops (e.g., what to do when status shows a failed story beyond the brief '--retry-failed' mention) remain implicit. | 4 / 5 |
Progressive Disclosure | The ~100-line body is well-organized into labeled sections with a clearly signaled one-level reference ('See docs/REFERENCE.md') and no nested references; no bundle files exist to verify further. Not 5 because parallel-session caveats and flag details are inlined and partially duplicated rather than split into a reference file, leaving minor organization gaps. | 4 / 5 |
Total | 15 / 20 Passed |