Delivery Orchestrator

This skill should be used when calling GitHub MCP tools — which tool does what, the method discriminators several of them require, oversized payload handling, and the pitfalls that produce confusing errors.

This skill should be used when the user asks to "triage issue", "classify ticket", "route jira", "analyze priority", "categorize issue", "determine complexity", "route to agents", or needs guidance on classifying, prioritizing, and routing Jira issues to appropriate agents and workflows.

This skill should be used when writing or reviewing a pull request title, description, or size — covering template resolution, PR size budgets, conventional titles, and what must never appear in a PR body.

This skill should be used when the user asks to "reason through", "think step by step", "analyze systematically", "debug with hypotheses", "root cause analysis", "chain of thought", "tree of thought", "documentation lookup", or needs structured reasoning frameworks with mandatory documentation retrieval before analysis.

This skill should be used when the user asks to "create a Confluence page", "write documentation", "write a TDD", "create an ADR", "write a runbook", or "publish to Confluence" — authoring and linking technical documentation in Confluence from Jira work.

This skill should be used when the user asks to "deploy with Harness", "set up Harness CD", "configure a canary/blue-green/rolling deployment", "add approval gates", or "roll back a deployment" — Kubernetes, Helm, Terraform, ECS, and serverless delivery via Harness.

This skill should be used when the user asks to "set up Harness CI", "build a CI pipeline", "configure test intelligence", or "cache/parallelize builds" — container-native build infrastructure on Harness.

This skill should be used when implementing Linear integrations using the official @linear/sdk — creating issues programmatically, advanced fetch/modify patterns, error handling, and type safety. Activates on "linear sdk", "@linear/sdk", "linear client", "linear typescript".

This skill should be used when implementing or debugging two-way sync between Linear issues and a git host — branches, pull requests, reviews, deploys — across GitHub and Harness Code. Activates on "linear github sync", "harness sync", "harness bridge", "linear vcs", "issue to pr sync", "magic words", "closing keywords".

The role catalogue and the descending-model rule for a multi-session fleet - which roles exist at each tier, what each owns, which model tier each runs on, and when a role should be its own session versus a subagent. Use when standing up a fleet, deciding who owns an activity, choosing a model for a spawned agent, writing a role brief, adding a second lane to one repository, or building the RACI for a run. Enforces that every spawned agent names its model - the orchestrator reasons, mid-tier executes, cheap models verify mechanically - and that a role which must heartbeat cannot be a subagent.

This skill should be used when choosing, diagnosing, or configuring how something authenticates to GitHub — personal access tokens, GitHub Apps, OAuth, SSH and deploy keys, GITHUB_TOKEN in Actions, OIDC, and the permission model each one carries.

This skill should be used when the user asks to "use the Harness MCP", "connect Harness to Jira", "manage Harness pipelines/repos/PRs from AI", or "sync Jira to Harness" — AI-driven Harness CD operations and Git/PR automation.

This skill should be used when working with GitHub merge queues and branch protection — queue semantics, required checks, strict status checks, auto-merge, and why a merge is blocked.

This skill should be used when the user asks to "create a PR", "open a pull request", "merge changes", "set up a git/PR workflow", or "request review" — creating, managing, and merging pull requests linked to Jira issues.

This skill should be used when working with stacked or dependent pull requests — resolving topology, retargeting and rebasing after a parent lands, splitting an oversized branch, and force-push safety.

This skill should be used when reading CI logs and diagnosing failures — fetching Actions logs through the GitHub MCP, classifying failures by cause, and distinguishing flaky tests from real regressions.

This skill should be used when implementing Linear OAuth 2.0, OAuth actor authorization, or file-storage authentication. Activates on "linear oauth", "linear auth", "actor token", "linear-actor-token", "file storage".

This skill should be used when the user asks to "enrich a task", "add task details", "analyze requirements", "add context to an issue", "expand a ticket", or "estimate complexity" — enriching Jira issues with requirements, dependencies, and technical context.

This skill should be used when delegating work to agents that run inside GitHub — Copilot cloud agent, custom agents, Copilot code review, and GitHub Agentic Workflows — including when delegation beats doing the work directly.

This skill should be used when building, registering, or operating an agent inside Linear — agent sessions, agent activities, delegation, Agent Skills, Loops, and Linear's built-in coding sessions. Activates on "linear agent", "agent session", "agent activity", "delegate issue", "linear aig", "agents in linear", "coding session", "linear loops".

One claim with ten sightings - a wrong predicate and a missing notification both produce a plausible, quiet, checkable-looking state, neither raises an error, and the only defence in both is a second reading through a different mechanism. In tooling that is the raw print through a different field; in coordination it is measuring the thing you wait on rather than waiting for news of it. Covers - ask what an artifact SITS ON and whether that moved, print the RAW STATE beside any predicate that decides whether to stop, a gate must name an input that turns it red, a destructive predicate needs its spare case proven by mutation, an absence needs the size of the space searched, a search-space cap manufactures false negatives while a result-set cap does not, check the exit code before reading the output, a true measurement stated as a durable property never expires and never stops being wrong, a count is only evidence when the thing counted is named, and a correction is itself a claim that names the set it ranges ov

The contract every session in a multi-session Claude Code fleet reads first. Use when starting, joining, or running a fleet of parallel sessions on one codebase - heartbeats and standby, addressing other sessions through a dispatch router, ASSERTED vs DELIVERED claim states, git discipline (one repo per lane, push-and-open-PR as one action, never rewrite shared history), and merge-readiness discipline (re-read the checks as the literal last action before a merge). Triggers on "start a fleet", "orchestration protocol", "lane", "heartbeat", "dispatch", "PR ready", "merge gate", "who owns this repo", or any question about how parallel sessions coordinate.

This skill should be used when writing or reviewing GraphQL queries against the Linear API — selecting fragments, navigating connection types, batching, and avoiding deprecated fields. Activates on "linear graphql", "linear query", "linear api", "linear schema".

This skill should be used when reading or writing GitHub Projects (v2) — the REST and GraphQL surfaces, field types and their IDs, item lifecycle, built-in workflows, and the token types each endpoint accepts.

This skill should be used when registering, verifying, or processing Linear webhooks — HMAC signatures, replay protection, idempotency, dead-letter queues. Activates on "linear webhook", "webhook signature", "Linear-Signature", "webhook secret".

This skill should be used when the user asks to "run a Harness pipeline", "trigger a deployment pipeline", "monitor a Harness execution", or "check pipeline status" — executing and watching Harness pipelines with Jira integration.

This skill should be used when the user asks to "configure Harness delegates", "set up Harness RBAC", "manage connectors/secrets/templates", "apply OPA policy as code", or "review audit logs" — Harness platform administration and governance.

This skill should be used when running or designing a code review process — adversarial verification, lens separation, evidence standards, severity mapping, and blackboard synthesis.

This skill should be used when writing, hardening, or optimizing GitHub Actions workflows — permissions, action pinning, script injection, caching, matrices, concurrency, and cost.

Detect and replace a hung session in a multi-session fleet without killing a working one. Use when a lane looks silent, stalled, unresponsive or hung, when deciding whether to replace or ping a session, when taking a fleet census, or when a heartbeat file has gone stale. Enforces the hard-won rule that message-absence is not work-absence - read the lane's own artifact (heartbeat file, branch, open PR, worktree) and quote what it showed before calling anything silent - plus the staleness thresholds, the three silence classes and their different remedies, and the replacement handover that lets a successor resume rather than restart.

This skill should be used when computing or interpreting repository delivery metrics — DORA four keys, lead time decomposition, change-failure hotspots, ownership drift, bus factor, and flake cost.

This skill should be used when building automation on GitHub Actions beyond CI — manual and external triggers, scheduled jobs, reusable workflows, composite actions, issue and PR bots, auto-merge, and Dependabot.

This skill should be used when the user asks to "apply clean architecture", "follow SOLID principles", "add dependency injection", "separate layers", "use hexagonal/onion architecture", "domain-driven design", or needs guidance on structuring code into clean, testable, decoupled layers.

This skill should be used when working with GitHub's issue data model — issue types, sub-issues, dependencies, labels, milestones, issue forms and templates, and the community health files that configure them.

This skill should be used when the user asks to "orchestrate jira", "work on issue", "complete jira ticket", "development workflow", "jira automation", "issue lifecycle", "work on story", "fix bug ticket", or needs guidance on coordinating development work through Jira with multi-agent orchestration patterns.

This skill should be used when uploading files to Linear, linking external URLs as attachments, or downloading existing attachments with auth. Activates on "linear attachment", "linear file upload", "fileUpload mutation", "attachmentLinkCreate".

This skill should be used when cutting a release — computing a semver bump from conventional commits, writing changelogs and release notes, sequencing breaking changes, and planning rollbacks.

This skill should be used when running many parallel coding agents against a Linear backlog — conductor/worker dispatch, exclusive claims, workspace isolation, retry and stall handling. Activates on "swarm", "parallel agents", "conductor", "fan out issues", "symphony", "worktree isolation", "delegation ledger".

This skill should be used when coordinating multi-agent GitHub delivery work — dispatching teams, holding a blackboard, enforcing merge gates, and deciding what to run in parallel versus in sequence.

This skill should be used when designing bulk operations against Linear — complexity budgets, exponential backoff, request pacing, and avoiding 429s. Activates on "linear rate limit", "linear 429", "linear throttle", "complexity budget".

This skill should be used when fetching large result sets from Linear — cursor pagination, filter DSL, sorting, and complexity budgeting. Activates on "linear pagination", "linear filter", "linear cursor", "linear filtering".