Spawn a thinking-partner via a local opencode run (GPT 5.6 Sol by default, other GPT and Claude models selectable), warm it up with "Hello 👋", and report the session id so follow-up consultations land in the same thread.
OpenCode 2 plugin development for this repo - the module layout under src/modules, the harness-neutral tool contract and defineModulePlugin adapter that serve both OpenCode and the Claude Code MCP bridge, credentials and /connect integrations, logging, testing, and live-loading a plugin. Load before adding or changing…
TypeScript conventions - naming, types and modeling, Zod validation, the NestJS API layer and DTOs, services and repositories, errors, async, queues, data access and security - with a monorepo addendum. Load together with develop-code-quality before writing, refactoring or reviewing TypeScript.
Conventions for dispatching local opencode and claude CLI runs as agentic workers through the oc tools - sync vs async, model selection, session continuity, thinking-partner pattern, multi-model fanout, agent profiles for dispatched workers, long-running worker resilience (progress journals, usage-limit resume with…
Practical Git and GitHub CLI workflow conventions for branching, staging, diffing, committing, pushing, PR titles and descriptions, and ignore hygiene, plus every remote GitHub operation through gh - reading pull requests and diffs, waiting for CI on a pull request, a commit or a workflow run with githubciwait…
Git worktree conventions for orchestrations that fan out work across multiple agents editing the same repo concurrently. Load before dispatching parallel write-mode agents.
Human-voice, clarity, and anti-AI-pattern conventions for drafting, auditing, and rewriting any prose a person will read, so it reads as written by a specific human and carries its information in no more words than it needs. Load for any prose task: emails, blog posts, essays, reports, bios, social posts, newsletters…
Conducts workflow runs: writes a Workflow-shaped script for a phase of work, starts it with wfstart, waits with wfwait, reads the result and decides the next step. Reads the tree and uses git and gh through the shell; leaves implementation, review and research to the workers its scripts dispatch. Session profile…
Non-interactive implementation worker dispatched by ocrun or the orch supervisor: edits the repository, runs checks, commits and pushes its branch. Session profile: claude --agent implementer, or the agent a command pins; never chosen for ad-hoc delegation.
Reviews one pull request through a review command and submits the GitHub review with gh. Loads review skills; runs no tests, makes no repository edits or git writes. Session profile: claude --agent pr-reviewer, or the agent a command pins; never chosen for ad-hoc delegation.
Read-only investigator and thinking partner for recon, diagnosis, pair-program and adversarial-consult dispatches. Reads code and skills, searches the web, writes only scratch files. Session profile: claude --agent researcher, or the agent a command pins; never chosen for ad-hoc delegation.
Read-only reviewer for orchestrated review rounds and free-form reviews. Reads and greps the tree, loads review skills, writes only scratch files such as its round file, runs no tests and no git writes; CI on the head commit is the test evidence. Session profile: claude --agent reviewer, or the agent a command pins…
Language-neutral code-quality rules for writing and reviewing code in any language - naming by intent, removing needless indirection, when to extract, safe direct assignment, redundant work, magic values, dead code and unread state, single source of truth, explicit tie-breaks, and code comments. Load together with the…
Vertical spacing (blank lines) in code - how blank lines partition functions, tests, classes and modules into role-based paragraphs - plus a safe procedure for a spacing pass across many files or a whole repository with parallel agents. Load when writing or reviewing code layout, when code reads as one unbroken block…
Language-agnostic debugging playbook for backend, infrastructure, runtime, and solver/pipeline failures. Covers symptom-to-cause layer walking, cheap-first bisection, falsifying your own diagnosis, escalating-complexity as a wrong-premise signal, adversarial pressure-testing, verifying the end state over a fix's local…
Procedure for refactoring, cleanup, simplification and non-formatting code-consistency work in any language - the evidence that proves a refactor preserved behavior (contract-shape diffs, golden outputs, behavior tests), recording owner rulings before a large refactor, running it across modules with parallel agents…
AWS CLI workflow for agents through session-scoped IAM Identity Center (SSO) profiles created by the awssession tools - starting a session that holds only the environments the work needs, the one human-approved SSO login, the per-command --profile rule, ending the session, and command recipes for ECS rollouts and…
Retrospective workflow for identifying tooling friction and proposing self-healing improvements, run inside the session that did the work or from a handoff over its durable data (task directory, run logs, transcripts).
★not rated 1 7d agoA46 tokens
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: