gh-aw-fleet

Guide the operator through deploying the fleet.json-declared workflow set to a tracked repository using the gh-aw-fleet CLI. Use this skill whenever the user says "deploy to <repo>", "run fleet deploy for X", "push fleet workflows to X", "sync workflows to X via fleet", "apply default profile to X", or asks to update agentic workflows on a specific tracked repo. Also trigger when the user names one of the tracked repos (in fleet.json or fleet.local.json) in the same breath as deploy/push/apply/ship verbs. The skill handles dry-run review, the --apply go-ahead gate, and gracefully recovers from gpg-signing failures by composing a manual-finish shell paste rather than bypassing commit signing. Do not trigger for template catalog refreshes (that's fleet-eval-templates), onboarding a brand-new repo (fleet-onboard-repo), bumping pin refs (fleet-upgrade-review), or editing fleet.json structure.

Audit pin health and upgrade readiness across the gh-aw-fleet tracked repositories, then guide targeted upgrades. Use this skill whenever the user says "audit the fleet", "which repos need upgrading", "check pin health", "run fleet upgrade audit", "upgrade all repos", "what's out of date", "run gh aw upgrade across the fleet", or asks about the health of deployed agentic workflows across multiple repos. Also trigger on phrases like "is <repo> up to date" or "any repos behind on their pins". The skill runs `fleet upgrade --audit --all`, presents a per-repo dashboard, and handles individual repo upgrades one at a time with dry-run review and gpg-failure handling. Do not trigger for deploying new workflows (fleet-deploy), for refreshing the upstream catalog (fleet-eval-templates), or for onboarding a new repo (fleet-onboard-repo).

Aggregate per-repo api-consumption-report output across the fleet and recommend where to look for cost concentration. Use this skill whenever the user says "what's our fleet spending", "where is the cost concentrated", "which workflows are burning credits", "consumption rollup", "budget review", "cost by profile", "cost by cost-center", "monthly fleet usage", "what's the top burner", "any spend hotspots", or asks about Copilot-credit usage across multiple repos. Also trigger when the user asks "should we cut anything from `<profile>`" or "is `<repo>` pulling its weight" — both questions resolve to a consumption rollup grouped on the relevant axis. The skill runs `gh-aw-fleet consumption` with the right temporal and group-by flags, reads the structured output, and frames the result as a budget conversation (where the spend is, what's driving it, what to consider next). Do not trigger for deploying or bumping workflows (that's `fleet-deploy` / `fleet-upgrade-review`), for evaluating which upstream workflows to

Refresh the upstream template catalog (templates.json) and evaluate new or changed workflows against fleet profiles. Use this skill whenever the user says "check for new agentic workflows", "what's new in gh-aw or agentics", "refresh the template catalog", "evaluate new workflows", "run template fetch", "any new workflows worth adding", or "what changed upstream". Also trigger when the user asks about the upstream template catalog in general — additions, removals, or fit assessments. The skill runs `template fetch`, reads the resulting JSON in-chat (no subprocess LLM needed — the catalog stores full frontmatter + body), and recommends which new workflows deserve profile inclusion, which are niche, and which are internal/test noise. Do not trigger for deploying workflows to a repo (that's fleet-deploy), for pin-health auditing (that's fleet-upgrade-review), or for onboarding a new repo (fleet-onboard-repo).

Add a new repository to the gh-aw-fleet tracked set and walk it through its first deploy. Use this skill whenever the user says "add repo X to fleet", "track X with gh-aw-fleet", "onboard X", "start managing X", "register X in the fleet", "include X in our fleet", or provides a repo spec that isn't already in fleet.json / fleet.local.json and wants it managed. Also trigger if the user says "deploy to X" but X is not yet tracked — pivot to this skill before attempting deploy. The skill edits fleet.local.json to register the repo (never fleet.json, which is the public example), asks for profile selection and per-repo customization (engine override, exclusions, extras), verifies via `go run . list`, and then hands off to fleet-deploy for the first PR. Do not trigger for already-tracked repos (that's fleet-deploy), for fleet-wide audits (fleet-upgrade-review), or catalog evaluation (fleet-eval-templates).

Scaffold a new fleet profile — add a `profiles.<name>` entry to fleet.json (or fleet.local.json) with a curated workflow list and pinned source refs, and optionally mirror it to `profiles/<name>.json`. Use this skill whenever the user says "build a profile", "create a profile", "scaffold a profile", "new profile", "add a profile to fleet.json", "make a profiles/<name>.json", "start a docs-plus / security-plus / quality-plus / community-plus profile", or proposes a named profile that doesn't exist yet. Also trigger when `fleet-eval-templates` has surfaced candidates the user now wants materialized into a profile definition. The skill reads `templates.json` for workflow metadata, proposes the profile JSON inline, writes on approval, and for the `default` profile mirrors both files to preserve the documented verbatim-match invariant. Do not trigger for evaluating upstream workflows (that's `fleet-eval-templates`), for deploying an existing profile to a repo (that's `fleet-deploy`), for rotating pins on already-d