Skip to content

BumpyClock/agent-templates / skills/productivity/writing-for-agents

writing-for-agents

Unverified

Create or revise agent instructions in skills, AGENTS.md, or CLAUDE.md.

Writing for Agents

Write instructions that change useful decisions. Do not prescribe a process for every task.

Checklist

  • State the purpose, activation condition, and completion condition.
  • Preserve user intent, project constraints, and authorization boundaries.
  • Keep non-obvious rules that prevent a concrete failure.
  • Remove generic reminders. Remove duplicate rules that another section or document already owns.
  • Before adding a rule, name the decision it changes and one case where it does not apply.
  • Revise the section that owns the decision. Do not append an overlapping exception.
  • Put conditional procedures behind links. State when to read each link.
  • Use project terms consistently. Keep valid domain vocabulary.
  • Prefer observable outcomes over required artifacts, agent counts, or fixed sequences.
  • Apply Concise wording to every edit. Cut words only where meaning stays exact.

Keep the root document small enough to show its decisions. A short, single-purpose document needs no extra files. Keep useful examples and scripts when replacing them would require repeated work.

Shared execution defaults

When changing shared execution defaults, pick the applicable cases below as acceptance criteria. Judge task outcomes, unnecessary work, authorization, and response usefulness. Do not judge prompt recitation or mechanical compliance. These cases define expected outcomes, not measured results.

TaskExpected outcome
Small, isolated editDirect execution without unnecessary delegation
Independent, substantial tasksBounded delegation when its benefit exceeds coordination cost
Request for feedbackAnalysis without file changes
Required tool unavailableAn alternative that preserves behavior and restrictions, or an explicit blocker
Behavior changeRelevant validation and documentation, subject to repository policy

References

After structural edits, validate links and metadata. Separate editorial judgment from measured model behavior.

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: 464kUpdated Claude CodeCodexVerifiedInstall
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: 464kUpdated 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: 318kUpdated Claude CodeCodexUnverifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 318kUpdated Claude CodeCodexUnverifiedInstall