Load this at the start of any first session in the claude-code-framework repo — when you (human or model) have zero context and need the right mental model before doing anything. Triggers include "where do I start?", "give me a tour of this repo", "how does this framework fit together?", "what should I read first?"…
Load before touching setup.sh or setup.ps1 in claude-code-framework, or when asked "does the PowerShell installer match the bash one?", "why does setup work on macOS but fail on Windows?", "where do I add a new setup prompt / CCF variable?", or when a Windows CI setup job fails while bash tests pass. Covers the bash…
Load when installing or uninstalling the claude-code-framework in a target repo, or answering how installation works — driving setup.sh / setup.ps1, choosing flags (--dry-run, --reset, --non-interactive / -DryRun, -Reset, -NonInteractive), setting CCF env vars, using the global /install-framework skill, or explaining…
Load this before claiming any claude-code-framework change is "done", before running or extending tests/run-all.sh, or when a check-.sh test fails and you need to know what it actually gates. Trigger on "run the tests", "why did check-placeholders/check-dogfood-drift/check-consistency fail", "is this change covered by…
Gives the agent up-to-date documentation and code examples for libraries and frameworks, fetched on demand so answers are not stuck at the model's training cutoff. Runs locally from the @upstash/context7-mcp npm package.
Instructions for BostonOrange/claude-code-framework, covering agents.md — claude code framework, project overview, tech stack, system structure and key conventions.
Instructions for BostonOrange/claude-code-framework, covering claude.md — claude code framework, project overview, tech stack, system structure and key conventions.
Install the Claude Code Framework into the current repo by cloning it from GitHub (no local clone path needed). Global/account-level skill. Distinct from the per-repo /setup, which refines an already-installed framework.
Add or update knowledge reference files for skills — domain knowledge (business rules, external systems), codebase inventories (entities, classes, APIs), or both. Use when documenting knowledge for AI context.
Automated readiness gate for tickets before implementation. Scans for TBDs, vague acceptance criteria, missing specs, unresolved questions, and integration gaps. Use before /develop to catch incomplete tickets early.
Orchestrate deployments to target environments. Handles merging PRs, deploying, posting notifications, and updating work items. Use when deploying to staging/UAT/production.
Full development cycle - fetches ticket from work item tracker, analyzes, implements, validates, fixes build errors, and creates PR. Use when developing a ticket end-to-end.
Create new tickets from business requirements or design documents. Analyzes requirements, scans codebase, identifies gaps and open questions, and drafts complete refined tickets.
Triage production errors from your monitoring system. Queries new errors, groups patterns, analyzes root cause with codebase context, and lets developer create tickets or dismiss. Supports unattended CI mode.
AI Software Factory — end-to-end pipeline from approved ticket to deployed test environment. Chains readiness gate → develop (factory mode) → PR → CI deploy. Reduces human touchpoints from 5+ to 2.
Fetch library/framework documentation via context7 or web search and persist it to the knowledge base. Automatically invoked when documentation is needed for implementation tasks.
On-demand precise impact analysis for a symbol or file before changing it. Spawns impact-analyzer to grep callers, classify the cascade (direct/indirect/test-only), and produce a confidence-scored report. Different from .claude/state/architecture.md (broad map) — this is the focused per-symbol drill-down.
Build or refresh a vector index of the codebase at .claude/state/codebase.db for semantic search via /search. Opt-in for large codebases (>50k LOC); small projects get sufficient awareness from .claude/state/architecture.md + Grep.
Plan → code → review → re-code feedback loop with persistent state across iterations. Spawns the review-coordinator, tracks findings across pushes, dedupes against prior iterations, and respects user "won't fix" decisions.
AI-powered merge conflict resolution. Reads both features' story docs to understand intent, resolves conflicts per file type, validates merged result. Works for any codebase.