Agent Claude Code
Runs mechanical tasks with no judgement involved — regenerating docs, bumping dependencies, running commands and reporting output, formatting. Use proactively for anything that has one obvious correct outcome.
Agent Claude Code
Writes the minimal implementation to make a specific failing test pass. Use only when a confirmed-failing test already exists. Never use to write tests or to plan.
Agent Claude Code
Read-only codebase reconnaissance. Use to find existing patterns, locate where something is implemented, or understand an unfamiliar module before planning or building. Use proactively before writing a spec.
Agent Claude Code
Reviews a complete PR diff against the spec as the final automated gate. Use in CI once a PR is open. Read-only, reports findings by severity.
Agent Claude Code
Turns an accepted ADR into a buildable spec folder. Use when an ADR exists in docs/adr/ and there is no spec folder yet, or when an existing spec needs re-planning. Do not use for implementation.
Agent Claude Code
Writes the human-readable QA summary when a PR slice is complete, before the PR is raised. Use after all tasks in a slice are done and the suite is green.
Agent Claude Code
Writes failing tests before implementation, and re-runs tests after changes to decide whether a task is done. Use at the start of every task and after every code change. Holds sign-off authority.
Hook Claude Code
Runs when you submit a prompt, before the agent sees it, executing route-rules.sh. From Ashutosh6393/Portfolio-MCP.
Hook Claude Code
Runs before the agent uses a tool for Write, Edit and Read tool calls, executing guard-paths.sh. From Ashutosh6393/Portfolio-MCP.
Hook Claude Code
Runs after a tool call finishes for Write and Edit tool calls, executing check-freshness.sh. From Ashutosh6393/Portfolio-MCP.
Hook Claude Code
Runs when the agent finishes a response, executing check-test-count.sh. From Ashutosh6393/Portfolio-MCP.
Settings file Claude Code
Agent settings declaring 4 hook events (UserPromptSubmit, PreToolUse, PostToolUse, Stop).
Skill Claude CodeCodex
Run the red-green TDD loop on the current slice of a feature spec. Use to start or resume building a spec, or when the user says continue working on a feature. Stops at the slice boundary for human review. Phase 3 of the build pipeline.
Skill Claude CodeCodex
Turn a design discussion into an Architecture Decision Record in docs/adr/. Use after a grilling session, or when the user points at a GitHub issue, Linear/Jira ticket, or describes a problem that needs a recorded decision. Phase 1 of the build pipeline.
Skill Claude CodeCodex
Turn an accepted ADR into a sliced technical spec under specs/. Use when an ADR exists and is ready to build, or when the user asks to plan, spec, or slice a feature. Stops at a human gate. Phase 2 of the build pipeline.
Instructions file CodexOpenCode
AGENTS.md instructions for Ashutosh6393/Portfolio-MCP, covering claude.md, read these, rules — load on demand, repository shape and the five rules that override convenience.
Instructions file
Claude Code instructions for Ashutosh6393/Portfolio-MCP, covering claude.md, read these, rules — load on demand, repository shape and the five rules that override convenience.