بنقرة واحدة
h2
يحتوي h2 على 17 من skills المجمعة من dcosson، مع تغطية مهنية على مستوى المستودع وصفحات skill داخل الموقع.
Skills في هذا المستودع
Schedule yourself to wake up every 5 minutes and shepherd an open PR — fix failing CI, address review-bot comments (push fixes for legit findings, resolve false positives, re-request review once addressed), reply to mechanical human comments with fixes, and flag architecture-level human comments back to the user. Exits the loop when the PR is green and review is clean. NEVER merges the PR.
Verify implemented plan docs against their actual code, add completion signoff metadata, and surface any gaps as new beads. Designed for scheduler/concierge agents coordinating multi-agent signoff after implementation is done.
Orchestrate the full planning lifecycle — from input assessment through architecture, plan writing, review cycles with convergence, and sign-off. Designed for concierge/scheduler agents coordinating multi-agent planning work.
Surgically trim verbosity and within-doc duplication from a plan doc without losing fidelity. Cuts review residue, restated rationale, illustrative code that prose+signature already covers, redundant adjacent sections, and prose-then-table restatements. Preserves all API/interface specs, acceptance criteria, URP/EO/AA commitments, decision rationale where alternatives matter, and review disposition tables. Run after plan-incorporate, before plan-work-completion-signoff.
Clean up a long-running work branch before landing into main — close beads, prune review docs, strip ceremonial comments, undo accidental cross-cutting refactors that weren't part of the task. Run after implementation completes, before opening the PR to main. Specific to the final land into the trunk, NOT for ordinary feature-to-feature merges along the way.
Independently review a plan doc and its test harness. Produces a review findings doc rated P0-P3. Run multiple times with different reviewers for independent perspectives.
Review the interfaces and seams between connected plan components to catch wiring omissions before implementation begins. Runs after individual plan reviews converge, before plan-to-beads. Supports vertical-slice (end-to-end request path) and horizontal-seam (cross-cutting contract) review modes.
End-to-end wiring audit of a component or application. Traces every user-facing entry point through the full request path, identifies what is wired vs stubbed/dead, and verifies test coverage. Works for any project type — web apps, APIs, CLIs, databases, mobile apps.
Produce a high-level architecture doc and — if the project is large enough — a plan index that lists sub-plans to be written later. Use after shaping to create the planning structure. Does NOT write sub-plans itself.
Generate a planning summary doc with review statistics, incorporation rates, convergence tracking, finding patterns, and document metrics across plan docs AND test harness docs. Supports multiple review rounds — each round gets its own section.
Review a code diff for a bead/feature. Produces a structured review doc with severity-rated findings. Run by reviewer agents after implementation is complete.
Decompose reviewed plan docs into implementation beads (epics + tasks) with correct dependencies. Reads plan docs, identifies components, and creates properly sized bd tasks.
Write a detailed plan doc and its companion test harness doc for a single component. Use after plan-architect has created the plan index.
Incorporate code review feedback — make code changes, commit with review trailers, update disposition table, and request re-review if needed. Mirrors plan-incorporate for code reviews.
Incorporate review feedback into a plan doc. Reads review files, updates the source doc, adds a disposition table tracking every finding, and deletes review files. Supports multiple review rounds — detects existing disposition tables and adds the next round.
Use this methodology when collaboratively shaping a solution with the user - iterating on problem definition (requirements) and solution options (shapes).
Adversarially stress-test a technical plan by verifying claims against real docs, running POC code, and updating the plan before you build.