Skip to content

hassan-mohiddin/freeflow / .deprecated/skill-tooling-pre-rewrite/skills/write-skill

write-skill

Verified

Use when creating or revising an agent skill.

references/development-loop.md

View on GitHub
# Skill Development Loop

Read this when moving a skill from an unevaluated draft through measured revision to a readiness decision.

## Draft

- Define the target behavior and failure pressure.
- Identify the earliest useful first-read prompt and one nearby case.
- Declare guaranteed prior context and optional dependencies.
- Write the smallest candidate that is complete on first read.
- Record it as Draft or Unverified in the owning eval artifact, report, task memory, tool result, or delivery response.
- Keep that authoring status out of the subject `SKILL.md`.
- Use static validation for structure only.

## Evaluate The Behavioral Claim

Read [Evaluate Skill](../../evaluate-skill/SKILL.md) and name the exact claim: activation, first-read behavior, nearby behavior, dependency composition, retained use, artifact outcome, or readiness.

Preserve an adequate existing case instead of creating a replacement for process appearance. Compare the exact declared variants under fixed prompt, fixture, tools, host, model, and thinking settings. Do not use one evidence class to imply another.

## Revise From Evidence

- Classify the failed boundary before editing.
- Find the description cue, instruction, placement, dependency, reference boundary, example, or stop condition that should have prevented it.
- Change one measured pressure point when possible.
- Keep unrelated instructions and criteria stable.
- Rerun the complete fixed case after the change.

Do not add broad prose, examples, references, or scripts until the failure shows why they are needed.

## Decide Readiness

Require evidence for every claimed behavior and host. Keep unavailable activation, composition, retained-use, artifact, or cross-host evidence explicit in the owning evidence surface, not in the subject skill body.

A strong candidate normally has:

- earliest-use activation evidence;
- first-read behavior with only guaranteed context;
- a nearby case that does not get hijacked;
- pressure cases where the baseline fails and the candidate improves behavior;
- exact dependency, multi-turn, artifact, and host evidence only when claimed.

The owner decides promotion after the evidence and residual gaps are visible.
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: 471kUpdated Claude CodeCodexVerifiedInstall
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: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
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 availableStars: 322kUpdated Claude CodeCodexVerifiedInstall