aaronjmars · Oct 16, 2026 · Oct 16, 2026
diff --git a/.claude/skills/aeon/references/secrets.md b/.claude/skills/aeon/references/secrets.md
@@ -122,6 +122,7 @@ Set with `gh variable set NAME "value"`.
 | `GLM_REASONING_EFFORT` | GLM gateway reasoning depth: `low` / `high` / `max` (default `high`) |
 | `STATE_BACKEND` | `file` (default) · `dual` · `issues` - where run state lives |
 | `HEALTH_ISSUES` | `0` disables the votable per-skill health Issues |
+| `AEON_RUNNER` | Runner label for every workflow job (default `ubuntu-24.04`), e.g. `blacksmith-2vcpu-ubuntu-2404` |
 | `NOTIFY_MIN_SEVERITY` | Suppress notifications below this level |
 | `NOTIFY_EMAIL_FROM` | Default `aeon@notifications.aeon.bot` - **must be a Resend-verified sender** |
 | `NOTIFY_EMAIL_SUBJECT_PREFIX` | Default `[Aeon]` |

diff --git a/.github/workflows/aeon.yml b/.github/workflows/aeon.yml
@@ -166,7 +166,7 @@ concurrency:
 jobs:
   run:
     if: github.event_name != 'issues' || github.event.label.name == 'ai-build'
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     timeout-minutes: 50
 
     steps:

diff --git a/.github/workflows/chain-runner.yml b/.github/workflows/chain-runner.yml
@@ -26,7 +26,7 @@ concurrency:
 
 jobs:
   run:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     # Each step waits up to wait_for_runs' cap (60 min) for one aeon.yml run, and
     # dev-loop chains five of them back to back (feature, review, repair,
     # re-review, proof). 60 here killed any chain mid-wait; 360 is the hosted

diff --git a/.github/workflows/ci-aeon-skill-sync.yml b/.github/workflows/ci-aeon-skill-sync.yml
@@ -36,7 +36,7 @@ concurrency:
 
 jobs:
   check:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: .claude/skills/aeon and plugin/skills/aeon are in sync

diff --git a/.github/workflows/ci-agents-md.yml b/.github/workflows/ci-agents-md.yml
@@ -38,7 +38,7 @@ concurrency:
 
 jobs:
   check:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: Check AGENTS.md is regenerated from STRATEGY.md

diff --git a/.github/workflows/ci-apps.yml b/.github/workflows/ci-apps.yml
@@ -51,7 +51,7 @@ concurrency:
 jobs:
   format:
     name: format - biome
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
@@ -66,7 +66,7 @@ jobs:
 
   dashboard:
     name: dashboard — typecheck, test, build
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     defaults:
       run:
         working-directory: apps/dashboard
@@ -94,7 +94,7 @@ jobs:
 
   cli:
     name: cli — typecheck
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
@@ -130,7 +130,7 @@ jobs:
 
   mcp-server:
     name: mcp-server — build
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     defaults:
       run:
         working-directory: apps/mcp-server
@@ -156,7 +156,7 @@ jobs:
 
   webhook:
     name: webhook — worker bundle
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     defaults:
       run:
         working-directory: apps/webhook

diff --git a/.github/workflows/ci-capabilities-parity.yml b/.github/workflows/ci-capabilities-parity.yml
@@ -42,7 +42,7 @@ concurrency:
 
 jobs:
   check:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: Check capabilities taxonomy parity

diff --git a/.github/workflows/ci-gate.yml b/.github/workflows/ci-gate.yml
@@ -11,7 +11,7 @@ concurrency:
 jobs:
   # The job id `gate` is the required status check on main - do not rename it.
   gate:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     timeout-minutes: 30
     steps:
       - name: Require all CI checks to pass

diff --git a/.github/workflows/ci-harness-cli.yml b/.github/workflows/ci-harness-cli.yml
@@ -61,7 +61,7 @@ concurrency:
 jobs:
   smoke:
     name: smoke (${{ matrix.harness }})
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     timeout-minutes: 15
     strategy:
       fail-fast: false

diff --git a/.github/workflows/ci-harnesses-json.yml b/.github/workflows/ci-harnesses-json.yml
@@ -40,7 +40,7 @@ concurrency:
 
 jobs:
   check:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: Regenerate harnesses.json + gateways.json and diff (ignoring generated timestamp)

diff --git a/.github/workflows/ci-packs-json.yml b/.github/workflows/ci-packs-json.yml
@@ -40,7 +40,7 @@ concurrency:
 
 jobs:
   check:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: Regenerate packs.json and diff (ignoring timestamp)

diff --git a/.github/workflows/ci-readme-catalog.yml b/.github/workflows/ci-readme-catalog.yml
@@ -60,7 +60,7 @@ concurrency:
 
 jobs:
   check:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: Validate the README skill tables against the first-party catalog

diff --git a/.github/workflows/ci-shellcheck.yml b/.github/workflows/ci-shellcheck.yml
@@ -49,7 +49,7 @@ concurrency:
 jobs:
   shellcheck:
     name: shellcheck (error-severity gate)
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: shellcheck the tracked shell surface

diff --git a/.github/workflows/ci-skill-category.yml b/.github/workflows/ci-skill-category.yml
@@ -39,7 +39,7 @@ concurrency:
 
 jobs:
   check:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: Every SKILL.md declares a valid category

diff --git a/.github/workflows/ci-skill-integrity.yml b/.github/workflows/ci-skill-integrity.yml
@@ -61,7 +61,7 @@ concurrency:
 
 jobs:
   verify:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       # A skill added without regenerating the lockfile would otherwise be

diff --git a/.github/workflows/ci-skill-packs.yml b/.github/workflows/ci-skill-packs.yml
@@ -56,7 +56,7 @@ concurrency:
 
 jobs:
   check:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: Validate the community pack registry and its docs table parity

diff --git a/.github/workflows/ci-skills-json.yml b/.github/workflows/ci-skills-json.yml
@@ -63,7 +63,7 @@ concurrency:
 
 jobs:
   check:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: Regenerate skills.json and diff (ignoring timestamp + per-skill sha/updated)

diff --git a/.github/workflows/ci-tests.yml b/.github/workflows/ci-tests.yml
@@ -58,9 +58,9 @@ jobs:
     # Blacksmith, not GitHub: ~50 short serial test steps, so single-core speed
     # decides. Median job 51s vs 64.5s on GitHub's public 4 vCPU runner (queue 7s
     # vs 2s), live bwrap checks pass; Blacksmith 4 vCPU was no faster (50s).
-    # Bench: runs 38823788873 + 38825096928. Back to ubuntu-24.04 if Blacksmith
-    # is down, since this job gates every PR through CI Gate.
-    runs-on: blacksmith-2vcpu-ubuntu-2404
+    # Bench: runs 38823788873 + 38825096928. The label comes from the
+    # AEON_RUNNER repo variable like every other job; unset = ubuntu-24.04.
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
       - name: Install deps

diff --git a/.github/workflows/messages.yml b/.github/workflows/messages.yml
@@ -42,7 +42,7 @@ concurrency:
 jobs:
   poll:
     if: github.event_name == 'schedule' || (github.event_name == 'workflow_dispatch' && inputs.message == '') || (github.event_name == 'repository_dispatch' && github.event.action == 'cron-tick')
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     timeout-minutes: 5
 
     steps:
@@ -289,7 +289,7 @@ jobs:
     # mirroring what the poller does inline. See scripts/telegram-route.sh.
     if: >-
       github.event_name == 'repository_dispatch' && contains(fromJSON('["telegram-command","telegram-callback","telegram-reply"]'), github.event.action)
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     timeout-minutes: 5
     steps:
       - name: Checkout repo
@@ -356,7 +356,7 @@ jobs:
     if: >-
       (github.event_name == 'repository_dispatch' && contains(fromJSON('["telegram-message","discord-message","slack-message"]'), github.event.action))
       || (github.event_name == 'workflow_dispatch' && inputs.message != '')
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     timeout-minutes: 30
 
     steps:

diff --git a/.github/workflows/scheduler.yml b/.github/workflows/scheduler.yml
@@ -37,7 +37,7 @@ concurrency:
 
 jobs:
   schedule:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     # 10 min: a big catch-up tick (many owed slots after a delivery gap) does one
     # gh call + a 1s spacer per dispatch and persists state once at the end, so a
     # 5-min cap could cancel it mid-way and lose every stamp (all re-fire).

diff --git a/.github/workflows/setup-commands.yml b/.github/workflows/setup-commands.yml
@@ -24,7 +24,7 @@ permissions:
 
 jobs:
   setup:
-    runs-on: ubuntu-24.04
+    runs-on: ${{ vars.AEON_RUNNER || 'ubuntu-24.04' }}
     timeout-minutes: 5
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7

diff --git a/plugin/skills/aeon/references/secrets.md b/plugin/skills/aeon/references/secrets.md
@@ -122,6 +122,7 @@ Set with `gh variable set NAME "value"`.
 | `GLM_REASONING_EFFORT` | GLM gateway reasoning depth: `low` / `high` / `max` (default `high`) |
 | `STATE_BACKEND` | `file` (default) · `dual` · `issues` - where run state lives |
 | `HEALTH_ISSUES` | `0` disables the votable per-skill health Issues |
+| `AEON_RUNNER` | Runner label for every workflow job (default `ubuntu-24.04`), e.g. `blacksmith-2vcpu-ubuntu-2404` |
 | `NOTIFY_MIN_SEVERITY` | Suppress notifications below this level |
 | `NOTIFY_EMAIL_FROM` | Default `aeon@notifications.aeon.bot` - **must be a Resend-verified sender** |
 | `NOTIFY_EMAIL_SUBJECT_PREFIX` | Default `[Aeon]` |