Skills Catalog
Complete catalog of all 51 pair skills organized by type, process, capability, workflow, and meta.
pair ships with 51 skills organized into four types: process (lifecycle orchestration), capability (atomic operations), workflow (delivery stages dispatched by the batch engine), and meta (navigation). Each skill is a SKILL.md file invoked as a slash command by your AI coding assistant.
Last updated: 2026-09-24. Source:
packages/knowledge-hub/dataset/.skills/
Meta Skills
| Skill | Command | Description |
|---|---|---|
| next | /pair-next | Determines the most relevant next action for your project by reading adoption files and PM tool state. |
| loop | /pair-loop | Unattended delivery loop: per iteration, selects eligible cards via pair-next, runs a dependency + mutex analysis, composes implement-batch for a mutex-safe parallel batch (or drives one card sequentially), hands the autonomy policy (until / prepare / merge) to the batch, which merges a review-approved card only when the merge gate allows it, and evaluates the stop predicate. |
Process Skills
Process skills orchestrate multi-step workflows by composing capability skills.
| Skill | Command | Description | Composes |
|---|---|---|---|
| brainstorm | /pair-process-brainstorm | Runs discovery in three fixed phases, grill interview, domain integration, backlog tree triage, on a free theme or an existing $root issue, landing an integrated Draft epic/story tree while leaving the PRD untouched. | /pair-capability-grill; /pair-capability-map-subdomains, /pair-capability-map-contexts, /pair-capability-record-decision (optional); /pair-process-plan-epics, /pair-process-plan-stories (conditional) |
| specify-prd | /pair-process-specify-prd | Creates or updates a Product Requirements Document through structured template analysis and iterative review, drafting a new PRD, or filling in a missing/outdated section. | , |
| bootstrap | /pair-process-bootstrap | Orchestrates full project setup, PRD verification, project categorization, checklist, standards, quality gates, PM tool, for a brand-new project, end to end. | /pair-process-specify-prd, /pair-capability-setup-pm, /pair-capability-record-decision, assess-* (optional) |
| plan-initiatives | /pair-process-plan-initiatives | Turns a PRD into prioritized (P0/P1/P2) strategic initiatives through structured analysis and validation. | /pair-capability-write-issue |
| plan-epics | /pair-process-plan-epics | Breaks a strategic initiative into epics, each delivering end-to-end value in 2-4 sprints, through structured analysis and validation. | /pair-capability-write-issue |
| plan-stories | /pair-process-plan-stories | Slices an epic into user stories via vertical slicing and INVEST validation, each sized for one sprint. | /pair-capability-write-issue |
| refine-story | /pair-process-refine-story | Refines a user story from Draft to Ready, the single Draft→Ready path (D24): phase 0 grill(sync), Given-When-Then acceptance criteria, map-subdomains/map-contexts scoped analysis, classify matrix, sprint readiness. | /pair-capability-grill, /pair-capability-write-issue; /pair-capability-map-subdomains, /pair-capability-map-contexts, /pair-capability-classify (optional) |
| plan-tasks | /pair-process-plan-tasks | Breaks a refined user story into implementation tasks, checklist, dependency graph, AC-coverage table, added to the story body; no separate task issues are created. | /pair-capability-write-issue |
| implement | /pair-process-implement | Implements a refined user story task-by-task, via a 5-step cycle per task (context, branch, implementation, quality, commit). | /pair-capability-verify-quality, /pair-capability-record-decision, /pair-capability-checkpoint, /pair-capability-publish-pr |
| review | /pair-process-review | Reviews a pull request through 6 sequential phases (5 review + optional merge with parent cascade), validation, technical review, adoption compliance, completeness, decision, to decide whether it merges. | /pair-capability-classify, /pair-capability-verify-quality, /pair-capability-verify-done, /pair-capability-record-decision, /pair-capability-analyze-debt, /pair-capability-assess-security (required); /pair-capability-verify-adoption, /pair-capability-assess-stack, /pair-capability-execute-manual-tests (optional) |
Capability Skills
Capability skills are atomic operations invoked independently or composed by process skills.
Domain Modeling Capabilities
| Skill | Command | Description |
|---|---|---|
| map-subdomains | /pair-capability-map-subdomains | Classifies business capabilities into DDD subdomains (core, supporting, generic) with a volatility rating, scoped to items just touched. |
| map-contexts | /pair-capability-map-contexts | Maps subdomains to DDD bounded contexts and derives the integration pattern between them (integration strength, socio-technical distance, volatility), scoped to items just touched. |
Reclassified from process skills: see Migration Notes.
Assessment Capabilities
All assess-* skills are output-only: they produce a proposal (rendered adoption content + target) plus a human-facing report and write no adoption files. Persistence of a proposal is delegated to /pair-capability-record-decision, the sole adoption writer, composed by the caller (a process skill or a human/agent). See ADR-009 (adr-009-assess-output-only) in the project adoption records.
When the opt-in package-local adoption override is active, assess-* load global rules plus the <pkg>/.pair/ of the packages the diff touches, report central↔local conflicts, and may PROPOSE activating the override, never activating it.
The only writes any of them perform are operational reports under .pair/working/reports/ (D14), assess-security's audit, assess-coupling's full-scope audit, and assess-cost's report-mode panel. Never adoption, never backlog items.
| Skill | Command | Description |
|---|---|---|
| assess-ai | /pair-capability-assess-ai | Evaluates and recommends AI development tools, coding assistants (Copilot, Claude Code, Cursor), agent frameworks, MCP integrations, models, when the choice is open. |
| assess-architecture | /pair-capability-assess-architecture | Evaluates and recommends an architecture pattern (layered, hexagonal, microservices, modular monolith, etc.) when the choice is open. |
| assess-cost | /pair-capability-assess-cost | Assesses the financial exposure of a change as a chromatic cost class (green/yellow/orange/red) against the KB cost-signal catalog resolved through the project's stack/architecture/infrastructure adoption, in two modes, review/refinement classification and period cost-monitoring reports. |
| assess-coupling | /pair-capability-assess-coupling | Assesses architecture coupling against the three-dimensional model (integration strength x socio-technical distance x volatility + balance rule). |
| assess-infrastructure | /pair-capability-assess-infrastructure | Evaluates and recommends infrastructure strategy, cloud provider, CI/CD pipeline, deployment pattern, IaC, environments, when the choice is open. |
| assess-methodology | /pair-capability-assess-methodology | Evaluates and recommends the development methodology (Scrum, Kanban, Lean, SAFe, hybrid), iterations, ceremonies, flow, when the choice is open. |
| assess-observability | /pair-capability-assess-observability | Evaluates and recommends the observability strategy, monitoring, logging, tracing, alerting, telemetry, when the choice is open. |
| assess-pm | /pair-capability-assess-pm | Evaluates and recommends which project management tool fits this project (Jira, Linear, GitHub Projects, etc.) when the choice is still open, proposes a choice, doesn't configure anything. |
| assess-stack | /pair-capability-assess-stack | Evaluates and recommends the tech stack, languages, frameworks, database, libraries, versions, when the choice is open: full evaluation at bootstrap, a single new dependency during implementation ('should we add Redis', 'is this new dependency consistent with our stack'), or composed by /pair-process-review when it finds an unlisted dependency in code. |
| assess-testing | /pair-capability-assess-testing | Evaluates and recommends the testing strategy, framework, pyramid distribution, coverage targets, TDD approach, when the choice is open. |
| assess-security | /pair-capability-assess-security | Assesses security posture against resolved rules (KB global + per-service + per-web-app + adoption project rules). |
Classification Capabilities
/pair-capability-classify applies the quality model (KB default + tech/risk-matrix.md delta) and owns no criteria of its own, downstream consumers (gate, review, /pair-next, automation) read the resulting matrix and tags without owning any criteria (D18).
| Skill | Command | Description |
|---|---|---|
| classify | /pair-capability-classify | Builds the classification matrix by applying the quality model (KB default + tech/risk-matrix.md adoption delta), from the story context in refinement, from the code/diff in review (confirm-or-raise, never lower). |
Analysis Capabilities
All analyze-* skills analyze and report: they propose no adoption decision and never block a PR. Distinct from assess-* (evaluate options + propose an adoption choice) and verify-* (conformance pass/fail). See ADL 2026-07-13-skill-naming-verb-taxonomy.md in the project decision log. Two of the three write nothing at all; analyze-delivery-metrics writes exactly one file (its own period panel under .pair/working/reports/metrics/) the same D14 operational-artifact exception assess-security's audit uses, and still no adoption content and no backlog item.
| Skill | Command | Description |
|---|---|---|
| analyze-code-quality | /pair-capability-analyze-code-quality | Produces an objective code quality report, complexity, size, coverage, duplication, maintainability score, for a repo or package, with actionable recommendations. |
| analyze-debt | /pair-capability-analyze-debt | Produces a technical debt report, categorized items (code, design, test, documentation, infrastructure), prioritized by impact x effort. |
| analyze-delivery-metrics | /pair-capability-analyze-delivery-metrics | Generates the delivery/AI-metrics retro panel for a period, bug resolution time, PR lead time, process adoption, computed from the adopted PM tool/code host against the delivery-metrics guideline (definitions, aggregation, low-sample rule, bug mapping) and written as ONE period-keyed panel under .pair/working/reports/metrics/: headline numbers + trend first, raw breakdowns collapsed, idempotent by period key. |
Verification Capabilities
| Skill | Command | Description |
|---|---|---|
| verify-adoption | /pair-capability-verify-adoption | Detects conformity/non-conformity of already-written code and configuration against the project's adopted decisions (architecture, tech-stack, infrastructure, etc.), for a given scope, a compliance audit, not a new-choice evaluation (that's /pair-capability-assess-stack and siblings). |
| verify-done | /pair-capability-verify-done | Checks a PR or work item against Definition of Done criteria (universal + project-specific), skipping already-passing criteria. |
| verify-quality | /pair-capability-verify-quality | Checks whether the codebase passes the quality gates the CI pre-merge gate would run for the item/PR risk tier, resolves the risk:* tag, applies the KB gate matrix (🟢 lint+type+build; +unit from 🟡; +integration/E2E on 🔴), and reports pass/fail per gate so local verification mirrors CI for the tier check set (tier-check parity, no over-checking on green work; the coverage-guardrail and secret-scan CI layers are NOT mirrored locally). |
Setup and Management Capabilities
| Skill | Command | Description |
|---|---|---|
| setup-pm | /pair-capability-setup-pm | Configures the project management tool, creates/wires it up, applies its implementation guide, updates way-of-working, records the decision, whether the tool was already chosen or picked here. |
| setup-gates | /pair-capability-setup-gates | Configures CI/CD quality gates, pipeline config, shared lint/format packages, husky hooks, the required deterministic secret-scanning job (gitleaks by default, R6.5/D24), and the required pair-review + pair-explicit-approval branch-protection checks that make the review unskippable, for the adopted tech stack. |
| setup-harness | /pair-capability-setup-harness | Configures an agent harness (pi, opencode, Claude Code) to execute pair's process: resolves the harness explicitly (never inferred), verifies its fitness against the project's declared harness/access-path requirements before writing anything, provisions access path / skill paths / context / model provider per the harness's guide, and confirms rather than rewrites on re-run. |
| manage-flags | /pair-capability-manage-flags | Manages a feature flag's lifecycle, creation, activation, deactivation, cleanup, per the feature-flags guidelines. |
Quality Assurance Capabilities
| Skill | Command | Description |
|---|---|---|
| design-manual-tests | /pair-capability-design-manual-tests | Designs a manual test suite, critical path files + suite README, for post-release validation, by analyzing released artifacts, architecture, and deployment targets. |
| execute-manual-tests | /pair-capability-execute-manual-tests | Executes a project's manual test suite against released or deployed artifacts, producing a structured PASS/FAIL report per critical path. |
Recording and Writing Capabilities
| Skill | Command | Description |
|---|---|---|
| record-decision | /pair-capability-record-decision | Records a decision, ADR (architectural), ADL (non-architectural), DDR (domain, 3-criteria gate), or an analysis-log entry ($type: analysis, written as Category: Analysis, decision-log/), and is the sole generic writer of adoption files, including the context map's decision-backed sections. |
| write-issue | /pair-capability-write-issue | Creates or updates an issue in the adopted PM tool from a type-specific template (bug, story, epic, etc.), including topical labels (e.g. tech-debt) for deliberate promotion; $mode: comment posts a comment on an existing item without touching its body (the non-destructive cross-link path). |
| estimate | /pair-capability-estimate | Estimates story points for a refined user story using the project's adopted estimation methodology; confirms rather than re-estimating if already sized. |
Discovery Capabilities
| Skill | Command | Description |
|---|---|---|
| grill | /pair-capability-grill | Interviews one question at a time, always with a recommendation, after exploring the KB/codebase first; sync mode drives systematic AI-human alignment on a story until shared understanding is confirmed. |
State and Handoff Capabilities
| Skill | Command | Description |
|---|---|---|
| checkpoint | /pair-capability-checkpoint | Writes and resumes a self-contained progress checkpoint (story, branch, tasks done, decisions, remaining todos) so work survives a context reset. |
| publish-pr | /pair-capability-publish-pr | Publishes a completed story branch as a pull request: realigns the generated mirrors from the local dataset (committing them separately when they drifted), runs the quality gate, creates or updates ONE PR from the pr-template (conditional sections filled only when pertinent), copies the story's classification tags, marks it ready-for-review, updates the board state, then enters the PR state flow, registers the required pair-review check as pending (merge blocked from t0) and dispatches the review to a clean-context subagent. |
Workflow Skills
Workflow skills are the delivery phases of the batch engine, one mutation boundary each, invoked by the engine (or by the in-session coordinator /pair-workflow-cycle) during a review ↔ remediation cycle, never cascade-suggested by /pair-next. Each writes one handoff under .pair/working/runs/<run>/<story>/.
| Skill | Command | Description |
|---|---|---|
| red-spec | /pair-workflow-red-spec | Stage 1 of the delivery workflow, preparation: owns the authoritative inventory (AC or finding → producer/grammar → inputs, representations, consumers → equivalence classes and interactions → executable evidence), the grouping of a remediation round and the executable acceptance contract, discriminating witnesses that are RED against the unfixed base, positive/already-correct controls that may pass, stable row ids, writing ONLY tests and consumed fixtures before any production edit, in remediation (review findings), repair (verifier rejection), revision (contract gap) and, only for a run already on the sealed-a0 path, initial modes. |
| red-verify | /pair-workflow-red-verify | Stage 2 of the delivery workflow, independent contract validation and deterministic seal in one execution: reproduces every witness and control of a prepared acceptance contract against the unfixed base, re-derives the inventory's classes and interactions from the authoritative producer, checks discriminance (a witness fails for the intended defect, a control may pass, a test-only fix fails on the injected regression), fixture consumption and fixScope, emits ALL concrete gaps it found in one typed rejection with stable row ids, and, when the contract is verified, runs red-snapshot.mjs seal in the same execution and returns the snapshot. |
| implement-phase | /pair-workflow-implement-phase | Stage 3 (initial) of the delivery workflow, implementation of one refined story inside its persistent worktree: on a fresh card above the base, writing the story's tests and code together test-first (no up-front contract) and fixing what an informal, unrecorded self-review against the AC and the review rules finds; on a run already carrying a sealed a0 contract, strictly above its RED snapshot, never touching a sealed test byte, following the project's implement process, verifying the tier-resolved quality gate, recording decisions, writing the checkpoint, and then publishing exactly one review-ready PR through the project's publish-pr skill in the same execution. |
| green-fix | /pair-workflow-green-fix | Stage 3 (remediation) of the delivery workflow, implementation against a sealed remediation contract: discovers the RED snapshot from Git (never from the prompt), changes production only inside fixScope, never touches a sealed test byte, re-runs the witnesses and the tier gate, commits GREEN above the seal, updates the PR, appends the cycle log and returns an evidence ledger. |
| review-phase | /pair-workflow-review-phase | Stage 4 of the delivery workflow, independent final verification of one exact PR head: the deterministic custody check first (red-snapshot.mjs verify / verify-chain: snapshot ancestry, sealed blobs byte-identical, no unlisted test change, fixScope respected, revisions as successors), then the fixer's evidence re-run, the approved witnesses and controls, every prior blocking finding with an explicit transition, the source delta and its directly affected consumers, cross-group interactions, the adopted risk-tier review passes (general via /pair-process-review, plus security / boundary / architecture lenses as the tier rises), unioned into one finding set with stable ids, one verdict and the head it reviewed. |
| contract-phase | /pair-workflow-contract-phase | Phase 0 of the delivery workflow: ensures the machine contract derived from a KB markdown template (verdict options, severities with explicit ranks, finding fields) is fresh, cache-by-hash via ensure-contract.mjs, regenerated only when the template changed, and returns it as the enum-locked return-value schema the reviewer is held to. |
| cycle | /pair-workflow-cycle | In-session coordinator for pair's delivery cycle: drives ONE card through implement → verify (a fresh card has no up-front contract) and, for each round of review findings, prepare → validate → green → verify, one stage per subagent, from an interactive Claude Code, Codex or pi (with pi-subagents) session, no dependency on Claude Code's Workflow tool. |
Skill Properties
All skills share these properties:
- Idempotent: Re-invocation is safe and resumes from the last incomplete step
- Composable: Process skills compose capability skills; capability skills can be invoked independently
- Adoption-aware: Skills read
.pair/adoption/files for project-specific decisions - Template-driven: Output follows templates from
.pair/knowledge/guidelines/
How to Use Skills
Skills are invoked as slash commands in your AI coding assistant:
For details on how skills are resolved and transformed, see the Skill Management Reference.
Package-Local Scoping
review (PR diff), assess-* (diff) and refine-story (packages the story references) load only the touched packages' <pkg>/.pair/ when the opt-in package-local adoption override is active, and may PROPOSE activating it, never activating it. Without the override nothing changes.
Plugin Skill
/pair-assistant is not part of the catalog above: it ships only in the optional Claude Code plugin, and the CLI never installs it into a project. See Claude Code plugin & /pair-assistant.