skills

Activate when a subagent completes work and needs continuation check. Activate when a task finishes to determine next steps or when detecting work patterns in user messages. Governs automatic work continuation and queue management.

Activate when user asks to commit, push changes, create a PR, open a pull request, or submit changes for review. Activate when process skill reaches commit or PR phase. Provides commit message formatting and PR structure. PRs default to dev branch, not main. Works with git-privacy skill.

Orchestrate the end-to-end development workflow with autonomous quality gates. Use when users ask to implement/run workflow phases, start or continue development, handle review/QA findings, or process actionable comments by routing through create-work-items -> plan-work-items -> run-work-items. Enforces TDD by default for implementation work and only pauses for genuine decisions.

Activate when user asks to release, bump version, cut a release, merge to main, or tag a version. Handles version bumping (semver), CHANGELOG updates, PR merging, git tagging, and GitHub release creation.

Create and classify typed work items (epic, story, feature, bug, finding, work-item) in the configured tracking backend. Use when users ask to create/capture work or when actionable findings/comments are discovered from review, PR feedback, QA, regressions, or defect reports.

Plan and prioritize existing typed work items with dependency and hierarchy validation. Use when users ask to sequence work, when new actionable findings/comments were captured, or when blockers/dependencies changed before execution.

Activate when user needs coordination, story breakdown, task delegation, or progress tracking. Activate when the pm skill is requested or work requires planning before implementation. PM coordinates specialists but does not implement.

Execute the next actionable work item from the configured backend with process quality gates and continuous state updates. Use when users ask to run queued work, continue after planning, or execute actionable findings/comments already captured in the backlog.

Activate when user asks for Test-Driven Development, test-first implementation, red-green-refactor, or enforcing tests before code. Use for feature work, bug fixes, and refactors; treat TDD as the default rule unless the user explicitly waives it.

Use when you need to reproduce or debug web UI flows (especially auth/OIDC) via the Agent Browser CLI, capture snapshots/screenshots, and extract redirect URLs and on-page errors deterministically. Includes install/setup guidance when the CLI is missing.

Activate when user needs AI/ML work - model integration, behavioral frameworks, intelligent automation. Activate when the ai-engineer skill is requested or work involves machine learning, agentic systems, or AI-driven features.

Activate when user needs architectural decisions, system design, technology selection, or design reviews. Activate when the architect skill is requested or work requires structural decisions. Provides design patterns and architectural guidance.

Activate when user needs API or backend testing - REST/GraphQL validation, integration tests, database verification. Activate when the backend-tester skill is requested or work requires backend quality assurance.

Activate when starting new work to check for established patterns. Activate when ensuring consistency with team standards or when promoting successful memory patterns. Searches and applies best practices before implementation.

Activate when performing git operations. MANDATORY by default - prevents direct commits to main/master, blocks destructive operations (force push, reset --hard). Enforces dev-first workflow where all changes go to dev before main. Assumes branch protection enabled unless disabled in settings.

Activate when user needs database work - schema design, query optimization, migrations, data modeling. Activate when the database-engineer skill is requested or work involves database design or performance tuning.

Activate when user asks to code, build, implement, create, fix bugs, refactor, or write software. Activate when the developer skill is requested. Provides implementation patterns and coding standards for hands-on development work.

Activate when user needs CI/CD or deployment work - pipeline design, deployment automation, release management. Activate when the devops-engineer skill is requested or work involves build systems or infrastructure automation.

Activate when creating any summary, report, or output file. Ensures files go to correct directories (summaries/, memory/, stories/, bugs/). Mirrors what summary-file-enforcement hook enforces.

Activate when performing git commits, creating pull requests, or any git operation. MANDATORY by default - prevents AI attribution (Co-Authored-By, "Generated with" footers). Does NOT block legitimate AI feature descriptions.

Plan and structure GitHub Issues work for epics, stories, bugs, findings, and work-items with consistent labels, priorities, and parent-child links. Use when users ask to set up GitHub planning/tracking, create typed issues, define issue hierarchies, or enforce issue taxonomy via gh CLI (preferred for token efficiency) with optional GitHub MCP fallback.

Retrieve, normalize, and report GitHub issue state continuously with prioritization and parent-child awareness. Use when users ask for GitHub issue status reports, backlog prioritization views, trend snapshots, or automated state tracking using gh CLI (preferred for token efficiency) with optional GitHub MCP fallback.

Activate when users need first-time ICA setup from scratch: install ICA CLI, configure sources, install baseline skills (including `ica-cli`), verify health, and hand off to day-2 CLI usage.

Activate when users ask how to use, configure, verify, upgrade, or troubleshoot the ICA CLI (`ica`) after initial setup, including source management and day-2 operations.

Activate when needing configuration values like git.privacy, autonomy.level, paths.*, team.default_reviewer. Retrieves ICA settings using dot notation from config hierarchy.

Activate when performing infrastructure, VM, container, or cloud operations. Ensures safety protocols are followed and blocks destructive operations by default. Mirrors agent-infrastructure-protection hook.

Universal MCP client for connecting to MCP servers with progressive disclosure. Use when you need to list MCP servers/tools or call an MCP tool against a server that is not already wired into the current agent runtime.

Internal shared helpers for ICA MCP tooling (client/proxy). Not intended to be invoked directly by users.

Activate when setting up MCP servers, resolving MCP tool availability, or configuring fallbacks for MCP-dependent features. Configures and troubleshoots MCP (Model Context Protocol) integrations.

Local stdio MCP proxy server that mirrors upstream MCP servers/tools and centralizes authentication (OAuth, headers/env). Register one server in your agent runtime, manage upstreams via .mcp.json and/or $ICA_HOME/mcp-servers.json.

Activate when user wants to save knowledge, search past decisions, or manage persistent memories. Handles architecture patterns, implementation logic, issues/fixes, and past implementations. Uses local SQLite + FTS5 + vector embeddings for fast hybrid search. Supports write, search, update, archive, and list operations.

Activate when multiple independent work items can execute concurrently. Activate when coordinating non-blocking task patterns in L3 autonomy mode. Manages parallel execution from selected tracking backend (config-driven), with .agent/queue/ fallback.

Activate when asked to auto-review and merge a PR. Runs a closed-loop workflow: subagent Stage 3 review -> fix findings -> re-review -> post ICA-REVIEW receipt -> merge (optional via workflow.auto_merge).

Activate when user needs test planning or QA strategy - test frameworks, quality metrics, test automation. Activate when the qa-engineer skill is requested or work requires a comprehensive quality assurance approach.

Activate when user needs requirements gathering - business analysis, specification development, user stories. Activate when the requirements-engineer skill is requested or work requires bridging business and technical understanding.

Activate when reviewing code, before committing, after committing, or before merging a PR. Activate when user asks to review, audit, check for security issues, or find regressions. Analyzes code for logic errors, regressions, edge cases, security issues, and test gaps. Fixes findings AUTOMATICALLY. Required at process skill quality gates.

Activate when user needs security work - vulnerability assessment, security architecture, compliance audits, penetration testing. Activate when the security-engineer skill is requested or work requires security review.

Activate when user wants to create a new skill or update an existing skill. Activate when extending capabilities with specialized knowledge, workflows, or tool integrations. Provides guidance for effective skill design.

Activate when users want to create or update an Agent Skill and want a Test-Driven Development approach. Defines a red-green-refactor workflow for SKILL.md authoring, trigger quality, validation, and iterative refinement.

Activate when user asks for improvement suggestions, refactoring ideas, or "what could be better". Analyzes code and provides realistic, context-aware proposals. Implements safe improvements AUTOMATICALLY. Separate from reviewer (which finds problems).