AlthafMjeelani/devbox-config

adversarial-verification

Proves guardrails, validators, permissions, hooks, and invariants with controlled pass-fail-pass checks. Use when implementing or reviewing enforcement mechanisms.

agent-cost-review

Use when recording real session token metrics, tool call counts, retries, and estimated costs per verified task.

api-contract-design

Use when designing or evolving REST, RPC, and event schemas with stable contracts, validation, versioning, compatibility, and error schemas.

api-design

Use when designing or reviewing HTTP APIs, service interfaces, events, schemas, request/response contracts, and integration boundaries.

architecture-review

Use when evaluating proposed designs against modularity, coupling, data flow, scalability, operational complexity, and existing idioms.

autoplan

Use when creating end-to-end implementation plans, feature plans, or agentic workflow plans before editing code.

browser

Use for web browsing, QA, and scraping - headless, CLI-driven, no visible window. Fetch page text, screenshot, click, fill forms, assert elements, run accessibility scans.

cache-and-reuse

Use when structuring instructions and context to maximize provider prompt caching and reusing verified findings across session steps.

ceo-review

Use when reviewing a plan, proposal, or change with a CEO/founder lens - problem framing, scope, user value, and strategic alternatives. Works at plan stage or on a diff.

ci-cd-and-release

Use when verifying formatting, linting, builds, test pipelines, migrations, deployment risks, feature flags, and rollback procedures.

code-review

Use when reviewing code, diffs, staged changes, or pull requests for correctness, regressions, maintainability, and missing tests.

codebase-navigation

Use when exploring unfamiliar repositories via package manifests, entry points, symbol searches, call sites, and test fixtures without reading entire files.

context-compaction

Use when preparing a loss-aware pre-compaction checkpoint before context resets or summarizing state to restore execution afterwards.

context-manager

Use when identifying the minimum relevant project or feature context, discovering entry points, and tracking authoritative working context across tasks.

crew

Use when orchestrating several agents in parallel - split a request into independent features, dispatch one crewmate per feature with the `crew` CLI, then stop and report status on request (not continuously) so ready branches can be reviewed and shipped.

critique

Use when providing an adversarial second opinion on a plan, diff, review, or recent analysis.

database-migration-safety

Use when reviewing and authoring database schema changes for table locks, backfills, transactions, reversibility, and deployment sequencing.

database-review

Use when reviewing schemas, migrations, queries, indexes, transactions, data retention, and database-related behavior.

debugging

Use when diagnosing bugs, failing tests, build errors, runtime failures, flaky behavior, or performance regressions.

decision-log

Use when recording significant architectural decisions, alternatives considered, trade-offs, rationale, and superseding decisions.

design-review

Use when reviewing UI/UX (planned or implemented) - user flow, design-system consistency, accessibility, state coverage, and AI-slop patterns. For building UI, use frontend-design instead.

documentation

Use when writing or updating README files, docs, examples, setup instructions, API notes, or developer-facing explanations.

documentation-sync

Use when updating affected technical documentation, README files, API specs, and runbooks alongside code changes to prevent stale claims.

effort-estimate

Use when estimating effort, timeline, cost, budget, phases, or rough quotation for client work. Defaults to INR unless another currency is specified.

eng-review

Use when reviewing an engineering approach (planned or implemented) - architecture, data flow, edge cases, failure modes, tests, and migration/operational risk.

explain

Use when explaining code, behavior, repository structure, or implementation details with file references.

feature-implementation

Use when implementing focused production changes following existing patterns, clear tests, safe error handling, and self-review diffs.

find-skills

Use when searching for public agent skills, evaluating whether a skill is safe to install, or deciding between existing and custom skills.

frontend-design

Use for generic frontend UI, UX, responsive layout, accessibility, and visual design improvements across web-facing projects.

gate-review

Use when the ship gate needs a structured code review - inspect the branch diff and output findings as a single JSON object classified auto_fix vs ask_user. Output only JSON, no prose.

git-and-pr-workflow

Use when managing git workflows, reviewing status and diffs, crafting conventional commits, and drafting pull requests safely.

git-commit

Use when creating git commits, drafting commit messages, branch names, or concise git summaries from staged or unstaged changes.

grill-me

Interviews the user to resolve requirements and decisions before action. Use only when the user explicitly asks to be grilled, interviewed, challenged, or questioned thoroughly.

grounding

Use when the agent might be hallucinating API signatures, library versions, or configuration syntax. Forces verification against actual code.

gsd-add-tests

Generate tests for a completed phase based on UAT criteria and implementation

gsd-ai-integration-phase

Generate an AI-SPEC.md design contract for phases that involve building AI systems.

gsd-audit-fix

Autonomous audit-to-fix pipeline — find issues, classify, fix, test, commit

gsd-audit-milestone

Audit milestone completion against original intent before archiving

gsd-audit-uat

Cross-phase audit of all outstanding UAT and verification items

gsd-autonomous

Run all remaining phases autonomously — discuss→plan→execute per phase