AI Agent Workflow Collection

Surface concrete in-task-collaboration protocols when the user describes an AI workflow informally — URL-as-entity-reference, durable PR-comment threads as session memory, "fit the protocol" basics. Trigger when the user names entities by colloquial label instead of stable URL, asks "how should I structure this for Claude", pastes a screenshot when a URL would do, or describes a multi-step Claude workflow without a durable handle. Apply tactics reactively, not as a checklist.

Verbalized Sampling (VS) protocol for deep intent exploration before planning. Use when starting ambiguous or complex tasks, when multiple interpretations exist, or when you need to explore diverse intent hypotheses and ask maximum clarifying questions before committing to an approach.

Review staged + unstaged changes and split them into one commit per logical change. Use whenever the user says "atomic commit", "commit my changes", "split this into commits", or has multiple unrelated edits sitting in the working tree — even if they don't say "atomic". Runs repo-native type-checker and linter before each commit and refuses to bundle unrelated changes.

Run the atomic-commit workflow on the current changes, then publish the resulting commits to the remote. Use whenever the user says \"commit and push\", \"ship these changes\", \"atomic commit and push\", \"publish my work\", or wants atomic commits delivered to origin in one step. Prefers git submit (git-branchless); falls back to a named branch + git push origin HEAD:refs/heads/ . Refuses force-push and direct push to protected branches without explicit authorization.

Publish a change-set as atomic GitHub issues/PRs — one issue/PR per logical change, never bundled. Use when the user says \"open a PR per change\", \"atomic PRs\", \"push issues and prs atomically\", \"individual PR for each change\", or wants a change-set delivered as separate PRs/issues. Asks per run which objects to create (PR-only vs issue+linked-PR). Detects whether the canonical repo is directly pushable via gh; if not, pushes branches to the fork and opens PRs targeting upstream with --head : . Requires gh auth; refuses force-push and direct push to protected branches without explicit authorization.

Compound a solved problem into a durable in-repo learning doc. Use when a verified non-trivial fix lands, the user says "compound this", "document this fix", or "remember this".

ODIN's compact-form conversation skill -- formal-logic English register with predicate claims, Hoare-triple framing, and ASCII shortened-English keywords. Trigger when user requests "axiom", "axiom-mode", "axiom-compact", or "compact form".

Use when the user wants to walk a complex design space round by round: map the decision tree, fire the current frontier as batched single-select questions, and loop until shared understanding is reached. Triggers: multi-fork decisions, ambiguous requirements, unresolved prerequisites, "batch ask me", "clarify the design space".

Demolish bloated code and re-derive it clean, cutting every surface the rebuilt contract does not name. Use for a repo-wide bloat sweep, when patching a subsystem has stopped paying, or when the user says "this whole module is bloated", "rewrite this properly", or "break it and rebuild".

Route a contributor to concrete, data-backed contribution opportunities by collecting repository signals, asking for the contributor's stated interest, and returning file-level recommendations with rationale and first action. Use when the user asks "where can I help", "what can I contribute", "find a good first issue", "find contribution opportunities", or "what should I work on".

Strip duplicate and conflicting directives across the system-prompt cascade family: the canonical baseline, the six output-style embeds, and the two external harness embeds.

Reduce concepts, duplication, and ceremony in internal code while touching nearby code. Use when working an existing path and you spot dead fields, redundant wrappers, or speculative abstractions; distinct from refactor-break-bw-compat (internal hygiene, not public API removal).

Commit working-tree changes and push to the current branch: no branch creation, no branch switch, no PR. Use when asked to push to the branch you are already on, including the default branch.

Coordinate context sweep before coding - gather relevant files, patterns, and tooling summaries. Use when preparing to implement a feature, fix, or refactor and need comprehensive architecture, pattern, tooling, and dependency context gathered first.

Design-by-Contract (DbC). Use when crossing public API boundaries, guarding complex state invariants, or hardening untrusted inputs and integration seams.

Ledger-first dedup of a skills/ or prompt-directory tree. Use when the user asks to dedup skills, find rules repeated across or within skill files, or check whether skills contradict each other.

Use when removing old code, migrating users to a new implementation, or deciding whether to maintain or sunset something.

Dependency-upgrade campaign — outdated scan, batch-by-severity, breaking-change remediation, lockfile audit. Use when CVEs require remediation, when a major upstream version lands, when stack compatibility forces a sweep, or on a scheduled (quarterly) hygiene cadence. CVE-driven bumps consume security audit findings as input.

Design-by-Contract (DbC) development - design contracts from requirements, then execute CREATE -> VERIFY -> TEST cycle. Use when implementing with formal preconditions, postconditions, and invariants across any language.

Ship the whole ask in one pass. Use when work would otherwise be split into phases, staged rollouts, follow-up PRs, or TODO-later remainders, or when the user says ship it now, no phases, or do it now.

Document decisions, not just code. Use when making architectural decisions, changing public APIs, shipping features, or capturing context for future engineers and agents.

Doubt-driven adversarial review. Use when correctness matters more than speed, the code is unfamiliar, stakes are high, a claim can't be checked by the type system or compiler, or verifying now is cheaper than debugging later.

Two-party working posture — user as director, agent as executor. Every fork, tradeoff, and taste choice is surfaced via batched AskUserQuestion with structural framing, a recommended default, and concrete previews when comparison is visual, so the human steers direction while the agent handles implementation. Eliminates the review-bottleneck (no giant diff to approve at the end — review is distributed across picks) and prevents codebase-understanding debt (the user remembers the architecture because they picked it). Use whenever the user invokes /duet, or says "work with me", "ask before", "check with me", "I want to decide", "don't assume", "human-in-the-loop", "co-author", "pair with me", "duet", or whenever a task clearly involves aesthetic, architectural, or irreversible strategic decisions — even without those exact words. Pair with the Duet output style to minimize cognitive load between picks.

Prove a decision space, state space, requirement set, or behavior surface is fully covered by enumerating every cell and classifying each as covered, gap, or deferred with an executed check. Use when the user says "exhaustive", "prove coverage", "did I miss any case", "enumerate the state space", or when a refactor or feature needs a completeness audit before it is called done.