hashgraph-online/awesome-codex-plugins / plugins/boshu2/agentops/skills/craft-goal
craft-goal
VerifiedDraft or lint a bounded long-running goal prompt with a finish line and hard limits. Use when: asked to craft or lint a goal (a /goal prompt); one change goes to Plan.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.9 KB |
| references/goal-prompt.md | markdown | 4.6 KB |
| scripts/validate.shexecutable | script | 273 B |
references/goal-prompt.md
View on GitHub# Goal prompt
Copy this prompt verbatim, replacing every angle-bracket field. Do not delete
the wave, hard-envelope, breaker and terminal-report sections.
```text
Goal outcome:
<larger caller-visible result>
Terminal acceptance and evidence:
1. Given <state>, when <event>, then <observable result> — <authoritative proof>
Domain terms:
- <term>: <one-line definition>; use only this word in beads, code and tests
Non-goals and authority:
- <excluded outcomes/mechanisms>
- Reads/writes/external/Git authority: <exact scope>
Bead graph:
- Tracker: <tracker with status, dependencies and notes>
- Root epic: <existing id or bounded bootstrap rule>
- First question: <falsifiable question the first wave tests>
- Initial experiments, if known: <bead → criterion/uncertainty>
- Record notes, scratch, evidence, verdict refs, and dependency/provenance links.
Experiment policy:
- Apply the `navigate` skill each wave: observe, pick, ratchet, checkpoint.
These policy lines bind with or without it.
- One build bead is one bounded experiment: native execution, or one RPI run
when the caller selects it. A decision bead resolves through the method
Navigate names for it, not through a build.
- An ordinary bead's checks and CI are its result. It gets one author-distinct
fresh validation only when the caller asks, a mistake cannot be cheaply undone
after it lands, or no deterministic check covers the change; a repair does
not start another.
- Select only work tied to an unmet criterion or named blocking uncertainty.
- Consume each verdict unchanged; useful progress needs evidence tied to an
unmet criterion or a blocking uncertainty, not digest/count movement alone.
- Distinguish pre-existing discovery from introduced regression using causal
evidence; unknown cause and recurrence require HOLD, not a design diagnosis.
- Classify discoveries as necessary-now, linked-follow-up, or HOLD/rescope.
Never downgrade a necessary-now discovery to linked-follow-up to obtain
completion.
- Retain evidence/provenance; revise or withdraw beliefs when evidence changes.
Wave envelope:
- <numeric experiment/concurrency/time/token/live-attempt limits>
Hard goal envelope:
- <numeric total experiment/time/token/live-attempt/compaction/surface limits>
- No artifact, repair, helper, retry, subject, compaction, or wave resets a total.
- Include helper and validation costs inside the allowance.
- Enforcing native controls and observable remaining allowance: <actual controls
and measurements; explicitly identify any unmeasured aggregate>.
- Objective text alone does not enforce a budget or a native pause.
Breaker and helper:
- Ordinary informative red may produce a materially different next experiment.
- HOLD implementation for causal review on: a repeated blocker, <threshold>
non-ratcheting results, oscillation between prior approaches, introduced
regression, unknown defect cause, recurrence, scope pressure, a requested
acceptance change, or an operator-reserved decision.
- A requested acceptance change or operator-reserved decision reports
NEEDS_OPERATOR at once, with no helper: only the caller can settle it.
- Any other HOLD incident gets at most one bounded fresh helper inside the
allowance, given acceptance, observations, failed approaches, exact evidence
and remaining allowance. Rewording the blocker or continuing the incident is
not a new incident.
- UNSTUCK names a materially different admissible experiment, its
discriminating check, and why it fits unchanged acceptance, authority and
bounds; the goal, not the helper, resumes, and a spent experiment bound stays
spent. ESCALATE, an unhelpful answer or no admissible experiment reports
NEEDS_OPERATOR; no more implementation or helper dispatch follows.
- Cancellation, an explicit refusal, or a spent hard time/cost/quota ceiling
stops at once with no helper. Cancellation or a spent ceiling reports
NOT_ACHIEVED with the exact gaps; a refusal reports NEEDS_OPERATOR naming
what was refused. A retry threshold alone is not a spent budget, and a
threshold for recording blocked is bookkeeping, not permission for more work.
Wave checkpoint:
- acceptance matrix; graph frontier; verdict/evidence summary;
- ratchets versus non-progress; measured remaining budgets; next thesis;
- observed native continuation/stop state, outstanding gaps, and helper use.
Never claim a native pause or aggregate enforcement based only on this text.
Terminal reports:
- ACHIEVED: every terminal criterion is proven.
- NOT_ACHIEVED: envelope or permitted search is exhausted, or the caller
cancels; report exact gaps.
- NEEDS_OPERATOR: judgment, rescope, refusal, or helper escalation; stop
implementation.
```
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |