Operating Autonomously
Runs a goal unattended with a disciplined progress / stuck / done loop, bounded retry-with-variation, and clear stop and escalation criteria. Use for long unattended runs, background jobs, or any "keep going until done" task. Wraps the decomposing-tasks -> sustained-execution spine and watches context-budgeting limits as a stop signal.
Decomposing Tasks
Turns a goal into a verifiable, risk-ordered plan with explicit success criteria before any work starts. Use at the start of any multi-step, ambiguous, or long-horizon task. Hands off to sustained-execution for long runs and verifying-reasoning to validate the plan.
Progressive Research
Finds external information without flooding context, a tiered search, triage, extract, corroborate, synthesize loop. Use when answering needs web search, docs, or unfamiliar APIs/libraries. For breaking a task into steps use decomposing-tasks; for raw window management use context-budgeting.
Auditing Skills
Scores and improves an existing Agent Skill against a spec + quality rubric, then proposes or applies concrete refactors (split monoliths, sharpen descriptions, fix portability). Use to review, grade, lint, or improve a skill, or to maintain a skill library at scale. Justifies each finding with verifying-reasoning and routes fixes back to authoring-skills.
Guarding Code Quality
Keeps agent-generated code correct, safe, and consistent with the codebase, match existing conventions, verify behavior with a real run/test, and self-review before claiming done. Use whenever writing or editing code. Calls verifying-reasoning to confirm a fix actually fixes, and reports honestly when checks fail.
Authoring Skills
Generates a new, spec-compliant, portable Agent Skill from a capability description, frontmatter, focused body, and the supporting file set, following progressive-disclosure and single-responsibility rules. Use when asked to create, scaffold, or package a skill. Sizes/splits via composing-skills and self-checks the output with verifying-reasoning, then hands to auditing-skills.
Composing Skills
Routes a task to the smallest set of skills that covers it and sequences them without context bloat, instead of one monolithic instruction. Use when a task spans multiple capabilities, when many skills are installed, or when deciding which skills to apply. Reads the skill registry to discover specialists and respects the tier/dependency ordering.
Skill Skeleton
Placeholder skill scaffold. Replace this entire frontmatter and body using authoring-skills before use. Not a functional skill on its own.
Sustained Execution
Keeps long, multi-session, or compaction-spanning tasks on track using an external task ledger, checkpoints, and resumable state. Use when work spans many steps/hours, may hit context limits, or must survive a reset. Pairs with decomposing-tasks (the plan), context-budgeting (the window), and operating-autonomously (unattended runs).
Verifying Reasoning
A fast, on-demand check that catches wrong answers before they ship, surface assumptions, seek disconfirming evidence, and validate the result against the question. Use before committing to a high-stakes conclusion, an irreversible action, or a final answer. Invoked as a quality gate by decomposing-tasks, progressive-research, guarding-code-quality, and the meta-skills.
Context Budgeting
Deliberately manages the context window to cut token cost and prevent quality decay on long runs. Use when a task is long-running, involves large files or many tool outputs, the context feels full, or cost matters. For finding external info efficiently use progressive-research; for surviving multi-session work use sustained-execution.