ai-engineering

Proactive governance advisor — checks standards, decisions, and quality trends during development. Always advisory, NEVER blocks. Three modes: `advise` (post-edit), `gate` (pre-dispatch), `drift` (on-demand decision audit). Trigger for 'governance check', 'advise on this change', 'check for drift', 'is this aligned with active decisions', 'shift-left advisory'. Not for blocking gates — use /ai-verify. Not for narrative code review — use /ai-review.

Designs motion, transitions, and micro-interactions for UI components: spring animations, gestures, easing, staggers, scroll-triggered and loading/skeleton motion. Trigger for 'animate this', 'add transitions', 'micro-interactions for', 'gesture design', 'swipe to dismiss', 'easing for this', 'review this motion for polish'. Not for design systems; use /ai-design instead. Not for visual art; use /ai-visual instead. Not for testing animation code; use /ai-test instead.

Delivers large multi-concern specs and backlog runs autonomously: decomposes specs into sub-specs (or normalizes work items into a backlog DAG), deep-plans with parallel agents, builds a dependency DAG, implements in waves, runs a single final quality loop with one bounded quality-remediation pass (verify+guard+review on full changeset), delivers via PR. Trigger for 'implement spec-NNN end to end', 'autopilot this', 'autonomous delivery', 'decompose and ship', 'run the backlog', 'execute these GitHub issues', 'process the sprint backlog'. Invocation is the approval gate. Not for small or single-concern tasks; use /ai-build instead. Not for ambiguous requirements; use /ai-brainstorm first.

Operates the project board (GitHub Projects v2 or Azure DevOps): discovers configuration after install (fields, state mappings, process templates) and syncs work-item state at lifecycle transitions. Trigger for 'set up the board', 'configure our ADO board', 'discover board fields', 'move this issue to in-review', 'update the board', 'mark as in progress', 'sync the work item state'. Two subcommands: `discover` (post-install configuration write) and `sync` (lifecycle state transitions). Auto-invoked via `sync` by /ai-brainstorm, /ai-build, and /ai-pr; fail-open. Not for backlog execution; use /ai-autopilot --backlog instead.

Forces rigorous design interrogation BEFORE any code: explores approaches, surfaces ambiguity, gathers evidence, produces an approved spec that becomes the contract for /ai-plan. Trigger for 'lets add X', 'how should we handle Y', 'whats the best approach', 'I am thinking about', 'what should we build for'. Not for existing approved specs; use /ai-plan instead. Not for execution; use /ai-build instead.

Cleans branches safely: switches to the default branch, prunes merged and squash-merged branches, syncs to remote, sweeps stale specs, rotates `.ai-engineering/runtime/` per retention policy. Trigger for 'tidy up', 'tidy branches', 'sync to main', 'delete old branches', 'start fresh', 'rotate runtime'. Auto-invoked by /ai-pr after merge. Not for committing changes; use /ai-commit instead. Not for code-level dead-code removal; use /ai-simplify instead.

Canonical implementation gateway: reads approved plan.md, resolves stack from manifest, deterministic-routes each task to its adapter, dispatches the build agent in an isolated worktree, runs TDD self-validation per task, then a single final quality loop with one bounded quality-remediation pass on the full changeset before /ai-pr. Trigger for 'go', 'start building', 'execute the plan', 'implement it', 'lets do this', 'build the plan', 'resume', 'continue'. Not without an approved plan; run /ai-plan first. Not for multi-concern specs needing decomposition; use /ai-autopilot instead. Not for a single function or subcomponent; use /ai-code.

Writes production code that satisfies stack-context standards on the first pass: interface-first design, backward-compatibility checks, lightweight self-review. Trigger for 'implement this', 'write the code for', 'add X to Y', 'build this function', 'make this work'. Not for tests; use /ai-test instead. Not for debugging; use /ai-debug instead. Not for refactoring; use /ai-simplify instead. Not for schema work; use /ai-schema instead. Not for executing an approved plan end-to-end; use /ai-build (the gateway).

Runs the governed commit pipeline: auto-branches from protected, stages selectively, formats and lints, scans for secrets, gates docs, composes a conventional message, pushes. Trigger for 'commit my changes', 'save my work', 'push this to remote', 'stage these files', 'ship it'. Not for opening a PR; use /ai-pr instead. Not for branch hygiene; use /ai-branch-cleanup instead.

Interviews the operator to produce a project-identity CONSTITUTION.md (Mission / Stakeholders / Vocabulary / Prohibitions / Compliance gates / Anti-goals / Boundaries / Escalation / Language / Lifecycle phase). Trigger for 'set up the constitution', 'define project identity', 'who is this project for', 'what does this project never do', 'amend the constitution'. Not for AI-behaviour rules — those live in CANONICAL.md / AGENTS.md. Not for spec governance; use /ai-governance instead.

Diagnoses broken behavior systematically with a 4-phase root-cause loop: test failures, runtime errors, crashes, regressions. Never patches symptoms. Trigger for 'it is not working', 'something broke', 'this used to work', 'I am getting an error', 'CI is failing', 'why is X happening'. Not for adding tests; use /ai-test instead. Not for security findings; use /ai-security instead.

Designs user interfaces and design systems: aesthetic direction, color palettes, typography, spatial composition, information architecture for web, mobile, CLI, and docs-heavy experiences. Trigger for 'design this page', 'create a design system', 'color palette for', 'typography for', 'UI for this feature'. Not for animation; use /ai-animation instead. Not for static visual art (posters, banners); use /ai-visual instead. Not for slide decks; use /ai-slides instead.

Manages the documentation lifecycle: CHANGELOG, README, solution-intent architecture docs, external docs portals, and documentation quality gates. Auto-invoked by /ai-pr. Trigger for 'update the changelog', 'the README is stale', 'document this feature', 'docs portal needs updating', 'did we document all changes'. Not for blog or pitch content; use /ai-prose instead. Not for marketing collateral; use /ai-marketing instead.

Files an upstream bug or improvement report against `arcasilesgroup/ai-engineering` (the framework repo) with strict seven-vector redaction, mandatory human confirmation, and an archived sanitized copy. Trigger for 'report this upstream', 'file an ai-engineering bug', 'this looks like a framework bug', 'tell anthropic / arcasiles about this'. Not for your own project's board (use /ai-issue); not for security disclosure (use the private channel listed in CONSTITUTION.md).

Explains code, concepts, patterns, and architecture with engineer-grade depth: 3-tier control (brief/standard/deep), ASCII diagrams, execution traces, anchored to real file:line references. Trigger for 'how does this work', 'why does it do X', 'trace through this', 'explain this pattern', 'walk me through'. Not for documentation artifacts; use /ai-prose or /ai-docs instead. Not for fixing code; use /ai-debug or /ai-build instead.

Codebase-only read-only research dispatcher. Thin wrapper around the ai-explore agent for architecture mapping, dependency tracing, pattern identification, and risk surfacing — dispatch for any question spanning >= 1 file, module, or import edge. Trigger for 'explore the codebase', 'where does X live', 'map this module', 'what depends on Y', 'trace this import chain'. Not for external evidence with citations; use /ai-research instead.

Validates framework compliance, ownership boundaries, risk acceptance lifecycle, and manifest integrity for regulated environments. Trigger for 'are quality gates enforced', 'who owns this file', 'formally accept a known risk', 'pre-release compliance check', 'governance report for auditors'. Not for code quality; use /ai-verify instead. Not for security scanning; use /ai-security instead — this validates governance process, not code content.

Audits an IDE end-to-end (instruction surface, hooks, skills, agents, installer wiring) using strict file-evidence — never assumptions. Trigger for 'audit IDE support', 'is Copilot wired up correctly', 'check Claude Code integration', 'are there orphaned hooks', 'verify IDE setup'. Accepts Claude Code, GitHub Copilot, Codex, Antigravity, or all. Not for code quality; use /ai-verify instead. Not for security scanning; use /ai-security instead.

Creates a project work-item (issue / task / story) on the user's configured board: routes by manifest `work_items.provider` (GitHub Projects v2 or Azure DevOps), composes title + body + labels, attaches to the active board, and confirms a clickable link. Trigger for 'open an issue', 'file a bug', 'create a task', 'add this to the backlog', 'log a work item'. Not for upstream framework bugs (use /ai-engineering-issue); not for board configuration (use /ai-board discover); not for committing code (use /ai-commit).

Extracts lessons from merged PR review feedback by analyzing what reviewers caught, identifying missed checks, and writing entries directly to LESSONS.md. Trigger for 'the AI keeps doing X wrong', 'learn from this PR', 'what patterns did reviewers catch', 'update our standards from feedback'. Not for in-session observation; use /ai-session-watch instead. Not for skill-level rewrites; use /ai-skill-improve instead.

Drives go-to-market execution and marketing content: blog posts for distribution, social crossposts, investor materials, market research, outreach campaigns, X/Twitter automation. Trigger for 'go to market', 'marketing plan', 'write a blog post to publish', 'crosspost to socials', 'market research for', 'investor deck', 'outreach campaign', 'content engine'. Not for internal docs; use /ai-docs instead. Not for sprint reviews or solution intent; use /ai-prose instead. Not for code explanations; use /ai-explain instead.

Audits MCP servers and skills on demand using LLM coherence analysis to catch capability drift and rug-pulls. Trigger for 'audit this skill', 'is this MCP safe', 'check coherence', 'detect rug-pull', 'snapshot baseline', 'mcp audit'. Three modes: scan (declared-vs-observed), audit-update (post-update diff), baseline set (anchor known-good). Not for runtime payload inspection; use prompt-injection-guard hook instead. Not for CVE scanning; use /ai-security instead.

Generates images, videos, and audio via AI models (fal-ai MCP): cheap iteration models, expensive production finals, cost-estimate before generation. Trigger for 'generate an image', 'create a thumbnail', 'make a voiceover', 'AI video', 'text to speech for'. Not for design composition; use /ai-visual instead. Not for animation specs; use /ai-animation instead.

Saves persistent technical discoveries (debugging insights, non-obvious behaviors, workarounds, integration gotchas) and searches them across sessions. Trigger for 'save this', 'note that', 'remember this finding', 'what did we find about', 'do we have notes on'. Rule of thumb: if it took more than 30 minutes to figure out, save it. Not for cross-session learning patterns; use /ai-session-watch or /ai-learn instead.

Onboards humans to a project: architecture tours, topic search, decision archaeology, structured new-team-member orientation. Read-only — never modifies code. Trigger for 'where does auth happen', 'what is the architecture here', 'history of this decision', 'onboard me to this repo', 'tour the codebase'. Not for agent session bootstrap; use /ai-start instead. Not for code-level explanation; use /ai-explain instead.

Generates, evolves, and validates CI/CD pipelines for GitHub Actions or Azure Pipelines, enforcing SHA pinning, timeouts, secret handling, and concurrency policy. Trigger for 'set up CI/CD', 'add a deployment pipeline', 'is this workflow secure', 'check workflow policy', 'add a security scan to CI'. Not for running pipelines; that is the CI system's job. Not for governance audits; use /ai-governance instead.

Decomposes an approved spec into a phased execution plan with bite-sized tasks, agent assignments, and gate criteria — the contract /ai-build executes. Trigger for 'break this down', 'create a plan', 'what tasks do we need', 'lets start implementing', 'scope changed re-plan'. Hard gate: user approves before /ai-build can run. Not for ambiguous requirements; use /ai-brainstorm instead. Not for execution; use /ai-build instead.

Documents production incidents, outages, degradations, and near-misses using the DERP format (Detection, Escalation, Recovery, Prevention) with targeted interview questions per phase. Trigger for 'we had an incident', 'write up the outage', 'something went wrong in prod', 'postmortem', 'near-miss analysis', 'incident report'. Not for customer support investigations; use /ai-support instead. Not for internal dev bugs; use /ai-debug instead.

Creates and updates pull requests with governance: runs the commit pipeline, enforces pre-push gates, generates structured PR body from spec, watches and fixes CI until merged. Trigger for 'open a PR', 'submit this for review', 'I am ready for review', 'merge this into main', 'draft PR', 'update the PR'. Not for commit-only flows; use /ai-commit instead. Not for narrative review; use /ai-review instead.

Optimizes prompts, system messages, and skill descriptions using explicit-over-implicit, show-do-not-tell, and rationale-embedding techniques. Trigger for 'this prompt is not working', 'optimize this skill description', 'improve triggering', 'rewrite this instruction', 'CSO-optimize'. Pass `--skill name` to optimize any skill's description field. Not for creating new skills; use /ai-scaffold instead. Not for evolving the entire skill body; use /ai-skill-improve instead.

Writes content (blog posts, pitch decks, sprint review summaries, architecture board reports, solution intent documents) with automatic audience targeting (developer/manager/executive). Trigger for 'write a blog post', 'pitch this', 'sprint review summary', 'architecture board doc', 'solution intent for'. Not for documentation artifacts like CHANGELOG or README; use /ai-docs instead. Not for marketing/social content; use /ai-marketing instead. Not for code-level explanations; use /ai-explain instead.

Measures AI system reliability over time by defining pass/fail criteria before implementation, running capability checks, and tracking regression via pass@k metrics. Trigger for 'how reliable is this', 'did my changes break anything', 'measure AI performance', 'define success criteria', 'eval this feature', 'check skill regression'. Not for code correctness; use /ai-test instead. Not for quality gates; use /ai-verify instead — evals measure AI task completion consistency.

External evidence with citations via a 4-tier escalation (local → free MCPs → web/Exa → NotebookLM autonomous deep research, launched first / harvested last). Every claim sourced [N] or marked [unsourced]; output ends with 3 cited recommended directions. Trigger for 'what does the state of the art say about', 'compare options for', 'find sources on', 'investigate this question', 'research this'. Use for questions whose answer lives OUTSIDE the codebase. Not for codebase exploration; use /ai-explore instead. Not for refactors; use /ai-simplify instead. Not for business-logic debugging; use /ai-debug instead.

Resolves git conflicts intent-aware: categorizes by type (lock files, migrations, generated, config, code), regenerates or merges per category, never blindly accepts. Trigger for 'I have conflicts', 'rebase failed', 'merge conflict', 'cherry-pick failed', 'unmerged paths'. Not for branch hygiene; use /ai-branch-cleanup instead. Not for committing the resolution; use /ai-commit instead.

Reviews code changes with human-quality judgment: PR reviews, file reviews, diff analysis, architecture feedback. Default mode runs the full specialist roster through 3 macro-agents; pass `--full` for one agent per specialist. Trigger for 'review this', 'give me feedback', 'look over my PR', 'any issues with this', 'is this merge-ready'. Not for evidence-backed gates; use /ai-verify instead. Not for narrative writing; use /ai-prose instead.

Creates new ai-engineering skills or agents end-to-end: scaffold, TDD pressure-test, optimize description, register in manifest, sync mirrors. Trigger for 'create a new skill', 'add a slash command', 'the framework needs a capability for', 'build a new agent', 'scaffold a skill for'. Not for evolving existing skills; use /ai-skill-improve instead. Not for description-only optimization; use /ai-prompt-tune instead.

Designs schemas, plans safe migrations with rollback scripts, optimizes slow queries with index recommendations, defines data retention and GDPR right-to-erasure policies. Supports PostgreSQL, MySQL, SQLite, MongoDB. Trigger for 'add a column', 'we need a migration', 'the query is slow', 'define a retention policy', 'GDPR compliance for data'. Not for application-layer ORMs without DB schema; use /ai-code instead. Not for security audits; use /ai-security instead. Not for infrastructure provisioning — no infra skill exists.

Runs security gates: SAST with OWASP/CWE mapping, dependency vulnerability scans, secret detection, SBOM generation for compliance, pre-release security verdict. Trigger for 'is this secure', 'audit dependencies', 'check for secrets', 'security report', 'is this package safe', 'compliance review'. Not for governance process; use /ai-governance instead. Not for runtime payload inspection; use prompt-injection-guard hook instead. Not for code quality metrics; use /ai-verify quality instead.

Observes session corrections, recoveries, and workflow patterns silently, then consolidates them into project-local observations on demand. Trigger for 'start observing', 'learn from this session', 'consolidate observations', 'review what was learned', 'observe this session'. Listening mode is passive; review mode extracts and writes. Not for cross-project learning; use /ai-learn instead. Not for skill rewrites; use /ai-skill-improve instead.

Consolidates the session-watch observation backlog on demand: runs /ai-session-watch --review, gates the result, and opens a draft chore PR for human review. Trigger for 'run the observation sweep', 'consolidate observations', 'session-watch sweep'. Never auto-merges, never auto-files work items, never runs unattended. Not for interactive review; use /ai-session-watch --review instead.