Heli-Harness

Plugin heli-harness in KJ-AIML/heli-harness. Commands read from the repo's manifest — 6 days ago.

Skills (31)

• audit

  Use when someone asks whether a change is correct, safe, or ready — read-only verification of a diff, PR, commit, or claimed fix.

• branch

  Use when creating, pushing, rebasing, merging, or opening a PR — branch and PR discipline without hard-coded repo policy.

• cloud-sync

  Use when any heli auth/ws/push/pull/sync/init command is about to run, when sync state or auto-push output appears in a session, or when the user asks about syncing workspace context across devices.

• concurrent-upgrade

  Use when an embedded compatibility workspace still has shared legacy task state and needs migration to the embedded concurrent task/session/lease layout; linked projects use resource-scoped authority instead.

• debug

  Use when a bug is confirmed but the cause is unknown — reproduce, isolate, root-cause, and explain before fixing.

• deps

  Use when changing package manifests or lockfiles — dependency add, remove, upgrade, replace, or pin protocol.

• design

  Use when design-before-code is needed for S2/S3 work, architecture-shaping changes, irreversible decisions, or API/data/security boundaries.

• engineering

  Use when non-trivial work needs risk tiering (S0–S3), done criteria, verification standards, or the smallest safe implementation path.

• evidence-gates

  Use when a diagnosis, failure transition, material decision, expensive retry, subsystem change, or completion claim needs current machine-readable evidence.

• feature

  Use when building a new capability or multi-step feature that needs a short brief, reviewable slices, and per-slice verification.

• fix-loop

  Use when tests fail, CI fails, fixes keep failing, or a wave of findings needs triage and one-cause-at-a-time closure.

• flow

  Use when the next protocol is unclear — lightweight router for ambiguous or mixed work. Route Heli-specific situational questions to heli-assistant.

• gh-write

  Use when creating or editing GitHub issues, PRs, comments, labels, milestones, or release notes with evidence-backed write discipline.

• heli-assistant

  Use when the user or agent needs situational help with Heli — what to do next, why a session/host/workspace is blocked, how to continue or migrate, whether a task/handoff is needed, or which Heli command/skill best fits the current goal.

• heli-audit

  Use when invoking the ergonomic /heli-audit entry point — compose broad workflow review with dependency, test-coverage, and impact lenses as triggered.

• heli-governance

  Use when working in a Heli linked project or embedded compatibility workspace to resolve layout, target/resource authority, policy, grants, host coverage, and evidence-backed completion.

• heli-help

  Use when the user asks what Heli can do, which command family exists, or how to find the right Heli workflow; route situational questions to heli-assistant.

• heli-impact

  Use when invoking the ergonomic /heli-impact entry point. Resolve the current Heli/repo context, then delegate blast-radius analysis to the canonical impact skill.

• heli-init

  Use when bootstrapping project facts/profile context for a Heli linked project or an embedded compatibility workspace.

• heli-install

  Use when installing, upgrading, linking, or migrating Heli. Prefer the public npm package plus heli setup/host install/link; embedded .heli-harness installs are compatibility/hermetic mode.

• heli-review

  Use when invoking the ergonomic /heli-review entry point. Resolve current Heli/repo context, then delegate focused verification to audit or broad review to workflow.

• heli-target

  Use when resolving or changing the active target/resource context. Prefer Heli CLI in linked projects; direct .heli-harness workspace files are embedded-compatibility only.

• heli-validate

  Use when invoking the ergonomic /heli-validate entry point. Resolve the repo's real verification context, then delegate safe command classification and execution to test-validation.

• impact

  Use when editing shared or high-use surfaces — pre-edit blast-radius analysis for callers, APIs, data, UI flows, jobs, and operations.

• incident

  Use when production is down, data is at risk, security is involved, or an urgent operational failure needs mitigate-first then root-cause protocol.

• release

  Use when preparing, cutting, publishing, deploying, tagging, or documenting a release; require pre-release gates and post-release proof that source, tag, registry/artifact, and fresh install agree.

• test-coverage

  Use when a bug fix, risky refactor, weak suite, coverage question, or regression-prevention need requires identifying missing, weak, flaky, or misleading tests.

• test-validation

  Use when bootstrapping or correcting a repo profile, or verifying that a profile's first verification command is safe, correct, and non-mutating.

• using-heli-skills

  Use when starting substantive Heli work to resolve layout, select the smallest workflow, and load specialist skills only when their triggers apply; use heli-assistant for situational Heli guidance.

• verify-premise

  Use when a claimed bug, review comment, or requested fix may be false — check the premise before implementing.

• workflow

  Use when a broad multi-file review, security/correctness sweep, or high-recall investigation needs find-broadly then refute-cold synthesis.

Hooks (3)

• SessionStart

  node "${CLAUDE_PLUGIN_ROOT}/hooks/heli-session-start.mjs"

• PreToolUse

  node "${CLAUDE_PLUGIN_ROOT}/hooks/heli-pre-tool-use.mjs"

• SessionEnd

  node "${CLAUDE_PLUGIN_ROOT}/hooks/heli-session-end.mjs"

Similar plugins