se-harness-copilot

Plugin se-harness-copilot from marketplace se-harness in rbhattarai/se-harness. Commands read from the repo's manifest — 28 days ago.

Skills (22)

• coding-discipline

  Karpathy-inspired coding discipline — think before coding, simplicity first, surgical changes, goal-driven execution — with real-world wrong/right examples. Use when implementing, reviewing, or fixing code that shows hidden assumptions, overcomplication, scope creep, or vague success criteria.

• context-injector

  Budgeted, progressive-disclosure memory context for every request. Use when deciding what project memory to load — never bulk-load memory files; follow the budget table and retrieve the rest on demand.

• harness-bootstrap

  Recommend and install the harness components matching this project's profile — methodology, stack plugins, MCP servers, observability — opt-in per item, recorded in the lockfile. `--reconfigure <topic>` revisits an earlier choice (observability, memory, methodology).

• harness-conventions

  Capture this repo's coding conventions and internal-library usage as per-area convention docs (angular, dotnet, e2e-playwright, db, ...) the agents load before writing code. Detect → draft → confirm each rule → write. Re-runnable; flags stale docs.

• harness-export

  Export the harness to another AI client — Copilot agent profiles, Cursor rules — with an honest capability report of what does and doesn't carry over.

• harness-goal

  Run the loop-engineering workflow for a user goal — grill, deep-dive, refine, story, implement, test, PR, deploy — with human gates at irreversible steps.

• harness-init

  Bootstrap the AI harness for this project — interview (new) or detect+confirm (existing), then generate profile, env, memory scaffold, org rules, and AGENTS.md/CLAUDE.md.

• harness-mem-codebase-memory

  Install, register, index and refresh the codebase-memory-mcp structural-memory driver for one repo or a whole workspace — optionally linking the indexed repos with its cross-repo mode. Confirms every third-party install first.

• harness-mem-codegraph

  Install, register, index and refresh the CodeGraph structural-memory driver (colbymchenry/codegraph, or codegraph-ai/CodeGraph) for one repo or a whole workspace. Confirms every third-party install and the telemetry choice first.

• harness-mem-graphify

  Build and maintain Graphify structural-memory indexes for one repo or an entire se-harness workspace — install, extract, cluster/label, export full-symbol HTML, and merge per-repo graphs into a workspace-level graph.

• harness-mem-sourcegraph

  Connect, verify and audit the Sourcegraph structural-memory driver (Cloud or self-hosted) — wire the MCP connection, check it works, and report which workspace repos the instance actually has indexed. Builds nothing; the instance does the indexing.

• harness-mem-wiki

  Build and refresh the workspace-level domain wiki from non-code sources — Jira, Confluence, SharePoint, NAS documents, NAS video — gated on each source actually being configured, synthesis-only, nothing ingested without confirmation.

• harness-methodology-openspec

  Install and initialize OpenSpec (Fission-AI/OpenSpec) for one repo or an entire se-harness workspace, gated on methodology:openspec actually being selected — setup only, /harness-goal drives the actual propose/archive lifecycle.

• harness-scan

  Scan an existing repo to detect stack, devops, cloud, topology, and org conventions; confirm with the user; merge into the harness profile and regenerate AGENTS.md/CLAUDE.md.

• harness-sync

  Detect drift between the project and its harness — stack changes, new recommendations, template updates, memory health — and refresh only generated content after confirmation.

• harness-update

  Update the se-harness plugin to the newest version (marketplace first, GitHub as fallback), then migrate this project's .harness/ data to match — without losing anything the project changed.

• memory-keeper

  Conventions for writing session/decision memory (tier 2) — daily logs, topics, scratchpad, typed causal links, promotion to the long-term index. Use at task end, after significant decisions/fixes, and when deciding where a learning belongs.

• requirement-grill

  Interrogate a goal until it is an unambiguous, testable requirement — scope, acceptance criteria, non-functionals, out-of-scope. Use at /harness-goal step 1 and whenever a request is vague enough that two reasonable engineers would build different things.

• stack-detector

  Evidence-based detection of a brownfield repo's stack, devops, cloud, and organization conventions. Use during /harness-scan — reason over collected evidence, never guess from a single package name, and always attach evidence + confidence to every claim.

• wiki-ingest

  Ingest an external source (Confluence page, Jira epic, URL, document) into the project's domain wiki — synthesize into pages, update the index, log the operation. Use when the user shares a source, when org conventions_url is set, or when /harness-goal finds an unread PRD.

• wiki-lint

  Health-check the domain wiki — contradictions, stale claims, orphan pages, index drift, missing links. Use periodically, after large ingests, or when the user asks whether the wiki is trustworthy.

• wiki-query

  Answer domain questions from the project wiki with citations — and file useful new syntheses back into it. Use during /harness-goal deep-dive (step 2) and whenever a question concerns domain knowledge rather than code.

Slash commands (14)

• /se-harness-copilot:harness-bootstrap

  Recommend and install the harness components matching this project's profile — methodology, stack plugins, MCP servers, observability — opt-in per item, recorded in the lockfile. `--reconfigure <topic>` revisits an earlier choice (observability, memory, methodology).

• /se-harness-copilot:harness-conventions

  Capture this repo's coding conventions and internal-library usage as per-area convention docs (angular, dotnet, e2e-playwright, db, ...) the agents load before writing code. Detect → draft → confirm each rule → write. Re-runnable; flags stale docs.

• /se-harness-copilot:harness-export

  Export the harness to another AI client — Copilot agent profiles, Cursor rules — with an honest capability report of what does and doesn't carry over.

• /se-harness-copilot:harness-goal

  Run the loop-engineering workflow for a user goal — grill, deep-dive, refine, story, implement, test, PR, deploy — with human gates at irreversible steps.

• /se-harness-copilot:harness-init

  Bootstrap the AI harness for this project — interview (new) or detect+confirm (existing), then generate profile, env, memory scaffold, org rules, and AGENTS.md/CLAUDE.md.

• /se-harness-copilot:harness-mem-codebase-memory

  Install, register, index and refresh the codebase-memory-mcp structural-memory driver for one repo or a whole workspace — optionally linking the indexed repos with its cross-repo mode. Confirms every third-party install first.

• /se-harness-copilot:harness-mem-codegraph

  Install, register, index and refresh the CodeGraph structural-memory driver (colbymchenry/codegraph, or codegraph-ai/CodeGraph) for one repo or a whole workspace. Confirms every third-party install and the telemetry choice first.

• /se-harness-copilot:harness-mem-graphify

  Build and maintain Graphify structural-memory indexes for one repo or an entire se-harness workspace — install, extract, cluster/label, export full-symbol HTML, and merge per-repo graphs into a workspace-level graph.

• /se-harness-copilot:harness-mem-sourcegraph

  Connect, verify and audit the Sourcegraph structural-memory driver (Cloud or self-hosted) — wire the MCP connection, check it works, and report which workspace repos the instance actually has indexed. Builds nothing; the instance does the indexing.

• /se-harness-copilot:harness-mem-wiki

  Build and refresh the workspace-level domain wiki from non-code sources — Jira, Confluence, SharePoint, NAS documents, NAS video — gated on each source actually being configured, synthesis-only, nothing ingested without confirmation.

• /se-harness-copilot:harness-methodology-openspec

  Install and initialize OpenSpec (Fission-AI/OpenSpec) for one repo or an entire se-harness workspace, gated on methodology:openspec actually being selected — setup only, /harness-goal drives the actual propose/archive lifecycle.

• /se-harness-copilot:harness-scan

  Scan an existing repo to detect stack, devops, cloud, topology, and org conventions; confirm with the user; merge into the harness profile and regenerate AGENTS.md/CLAUDE.md.

• /se-harness-copilot:harness-sync

  Detect drift between the project and its harness — stack changes, new recommendations, template updates, memory health — and refresh only generated content after confirmation.

• /se-harness-copilot:harness-update

  Update the se-harness plugin to the newest version (marketplace first, GitHub as fallback), then migrate this project's .harness/ data to match — without losing anything the project changed.

Subagents (13)

• architect

  Designs architecture for an approved requirement — impact analysis, component/data design, ADR. Use after a REQ is approved and before implementation is planned. Read-mostly; writes only design docs.

• db-engineer

  Schema changes, migrations, and query work for an approved requirement. Use during goal-loop step 5 when the design touches the data model.

• e2e-generator

  Generates Playwright specs from the e2e plan, exploring the live app via Playwright MCP for real selectors. Second of the planner→generator→healer chain.

• e2e-healer

  Diagnoses and fixes failing/flaky Playwright tests — selector drift, timing, data issues — or flags real app bugs. Third of the planner→generator→healer chain; also used in maintenance.

• e2e-planner

  Converts tracker test cases into an executable e2e test plan (scenarios, selectors strategy, data setup) for the Playwright generator. First of the planner→generator→healer chain.

• implementer-backend

  Implements backend/service-layer changes for one task of an approved requirement, with unit tests, in an isolated worktree. Use during goal-loop step 5.

• implementer-frontend

  Implements UI changes for one task of an approved requirement, with component tests, in an isolated worktree. Use during goal-loop step 5.

• integration-tester

  Writes and runs integration tests across component/service boundaries touched by the requirement — DB, queues, HTTP contracts. Use after unit testing, before e2e.

• release-manager

  Prepares releases — version bump, release notes from merged REQs/stories, deploy checklist. Runs ONLY after the human deploy gate (goal-loop step 9). Never deploys without explicit approval in the current session.

• reviewer

  Read-only design review of a REQ's merged diff — Karpathy discipline (scope, simplicity), clean code, and SOLID/design-pattern smells the diff itself introduced. Use in goal-loop step 5, after the tests pass and before the PR gate.

• story-writer

  Converts an approved requirement + design into a Jira user story and XRay/Zephyr test cases via the Atlassian MCP. Use at goal-loop step 4.

• unit-tester

  Strengthens unit-test coverage for changed code — gap analysis, edge cases, regression tests. Use after implementers finish, before integration testing.

• workspace-orchestrator

  Coordinates a multi-component /harness-goal task across a workspace's repos — reads workspace.yaml and REQ-NNN/workspace-plan.md, dispatches rows in dependency order, crosses into sibling repos directly when a row lives outside the current one, and keeps the plan's status current. Use at goal-loop s

Similar plugins