Output Skill — Complete Output Enforcement

Purpose

Prevents the agent from cutting corners during long execution runs. As context fills up, agents tend to produce lazy output: placeholder comments, skipped sections, TODO stubs, abbreviated code. This skill enforces complete output.

Rules (non-negotiable)

Never output these

• // TODO: implement this
• // ... rest of the code
• /* placeholder */
• // similar to above
• // add more as needed
• // etc.
• // implementation left as exercise
• Any comment that substitutes for actual code

Every function must be complete

• No function stubs with empty bodies
• No functions that say throw new Error('Not implemented')
• If a function is in the plan, write it fully. If it's not in the plan, don't create it.

Every file must be complete

• No partial files that "continue in the next task"
• Each file must be syntactically valid and importable
• All imports must resolve to actual modules

Every test must assert something meaningful

• No tests with expect(true).toBe(true)
• No tests that only check "does not throw"
• Every test must verify actual behavior

When this fires

This skill is auto-invoked on EVERY implementation task, alongside coding-standards. It serves as a quality floor that catches the most common agent failure mode: producing incomplete output when context gets long.

If the agent feels the urge to abbreviate

That is a context rot signal. Instead of abbreviating, the task should be split into smaller pieces, or the agent should report NEEDS_CONTEXT to the orchestrator.