Skip to content

MVPavan/via / .claude/skills/authoring-for-agents

authoring-for-agents

Verified

Use when creating or changing agent instructions or their activation behavior. Skip edits that only correct spelling or formatting.

Authoring for Agents

Write instructions for a named failure, a task-specific contract, or a repository fact the agent cannot infer reliably. Remove generic advice that adds no behavior.

Choose the surface

NeedOwner
Policy needed on every taskAGENTS.md
Convention needed for particular workProject document with a conditional AGENTS.md pointer
Repeated specialist workflowSkill
Explicit human entrypointManual skill or thin alias
Substantial material used by one branchConditional reference
Deterministic enforcementExisting hook, check, or script where authorized

Search for an existing owner before adding another. Keep shared semantics provider-neutral; runtime configuration owns models, effort, permissions and integration mechanics. Preserve scope restrictions on which files may change.

Write

  • Put the activation boundary first: positive cases and the nearest misleading case to skip. A description should select the skill, not duplicate its steps.
  • Inline the common contract; link substantial optional material at the branch that needs it. Moving text behind an unconditional read does not save context.
  • State outputs, material invariants, authority boundaries and stop conditions. Use exact recipes for fragile operations and judgment for routine choices.
  • Prefer one effective instruction over repeated prohibitions or ritual steps. Treat wording techniques as hypotheses to evaluate, not psychological laws.

For structural choices, consult references/writing-principles.md; for skill metadata, aliases and sidecars, consult references/skill-anatomy.md.

Verify

Inspect the diff for lost contracts and conflicting callers. For descriptions or routing changes, check a positive case and a nearby trivial/negative case through the actual discovery surface when available. Scale fresh-context and pressure trials to behavioral risk; methods are in references/testing-docs.md.

Structural checks establish packaging, not model behavior. Report unavailable behavioral checks as a limit; do not manufacture evidence or bypass runtime restrictions to run them. Revise the smallest instruction implicated by a failed case, then recheck that case and its affected neighbors.

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: 468kUpdated 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: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated 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: 321kUpdated Claude CodeCodexVerifiedInstall