Skip to content

Plugin System Design

Design document for the Gas Town plugin system. Written 2026-01-11, crew/george session.

Gas Town needs extensible, project-specific automation that runs during Deacon patrol cycles. The immediate use case is rebuilding stale binaries (gt, bd, wv), but the pattern generalizes to any periodic maintenance task.

Current state:

  • Plugin infrastructure exists conceptually (patrol step mentions it)
  • ~/gt/plugins/ directory exists with README
  • No actual plugins in production use
  • No formalized execution model

Reality is truth. State is derived.

Plugin state (last run, run count, results) lives on the ledger as wisps, not in shadow state files. Gate evaluation queries the ledger directly.

Agent decides. Go transports.

The Deacon (agent) evaluates gates and decides whether to dispatch. Go code provides transport (gt dog dispatch) but doesn’t make decisions.

LayerPlugin Analog
Moleculeplugin.md - work template with TOML frontmatter
EphemeralPlugin-run wisps - high-volume, digestible
ObservablePlugin runs appear in bd activity feed
WorkflowGate → Dispatch → Execute → Record → Digest

~/gt/
├── plugins/ # Town-level plugins (universal)
│ └── README.md
├── gastown/
│ └── plugins/ # Rig-level plugins
│ └── rebuild-gt/
│ └── plugin.md
├── beads/
│ └── plugins/
│ └── rebuild-bd/
│ └── plugin.md
└── wyvern/
└── plugins/
└── rebuild-wv/
└── plugin.md

Town-level (~/gt/plugins/): Universal plugins that apply everywhere. Rig-level (<rig>/plugins/): Project-specific plugins.

The Deacon scans both locations during patrol.

Key insight: Plugin execution should not block Deacon patrol.

Dogs are reusable workers designed for infrastructure tasks. Plugin execution is dispatched to dogs:

Deacon Patrol Dog Worker
───────────────── ─────────────────
1. Scan plugins
2. Evaluate gates
3. For open gates:
└─ gt dog dispatch plugin ──→ 4. Execute plugin
(non-blocking) 5. Create result wisp
6. Send DOG_DONE
4. Continue patrol
...
5. Process DOG_DONE ←── (next cycle)

Benefits:

  • Deacon stays responsive
  • Multiple plugins can run concurrently (different dogs)
  • Plugin failures don’t stall patrol
  • Consistent with Dogs’ purpose (infrastructure work)

Each plugin run creates a wisp:

Terminal window
bd wisp create \
--label type:plugin-run \
--label plugin:rebuild-gt \
--label rig:gastown \
--label result:success \
--body "Rebuilt gt: abc123 → def456 (5 commits)"

Gate evaluation queries wisps instead of state files:

Terminal window
# Cooldown check: any runs in last hour?
bd list --type=wisp --label=plugin:rebuild-gt --since=1h --limit=1

Derived state (no state.json needed):

QueryCommand
Last run timebd list --label=plugin:X --limit=1 --json
Run countbd list --label=plugin:X --json | jq length
Last resultParse result: label from latest wisp
Failure rateCount result:failure vs total

Like cost digests, plugin wisps accumulate and get squashed daily:

Terminal window
gt plugin digest --yesterday

Creates: Plugin Digest 2026-01-10 bead with summary Deletes: Individual plugin-run wisps from that day

This keeps the ledger clean while preserving audit history.


rebuild-gt/
└── plugin.md # Definition with TOML frontmatter
+++
name = "rebuild-gt"
description = "Rebuild stale gt binary from source"
version = 1
[gate]
type = "cooldown"
duration = "1h"
[tracking]
labels = ["plugin:rebuild-gt", "rig:gastown", "category:maintenance"]
digest = true
[execution]
timeout = "5m"
notify_on_failure = true
+++
# Rebuild gt Binary
Instructions for the dog worker to execute...
# Required
name = "string" # Unique plugin identifier
description = "string" # Human-readable description
version = 1 # Schema version (for future evolution)
[gate]
type = "cooldown|cron|condition|event|manual"
# Type-specific fields:
duration = "1h" # For cooldown
schedule = "0 9 * * *" # For cron
check = "gt stale -q" # For condition (exit 0 = run)
on = "startup" # For event
[tracking]
labels = ["label:value", ...] # Labels for execution wisps
digest = true|false # Include in daily digest
[execution]
timeout = "5m" # Max execution time
notify_on_failure = true # Escalate on failure
severity = "low" # Escalation severity if failed
TypeConfigBehavior
cooldownduration = "1h"Query wisps, run if none in window
cronschedule = "0 9 * * *"Run on cron schedule
conditioncheck = "cmd"Run check command, run if exit 0
eventon = "startup"Run on Deacon startup
manual(no gate section)Never auto-run, dispatch explicitly

The markdown body after the frontmatter contains agent-executable instructions. The dog worker reads and executes these steps.

Standard sections:

  • Detection: Check if action is needed
  • Action: The actual work
  • Record Result: Create the execution wisp
  • Notification: On success/failure

Current escalation is ad-hoc “mail Mayor”. Issues:

  • Mayor gets backlogged easily
  • No severity differentiation
  • No alternative channels (email, SMS, etc.)
  • No tracking of stale escalations

New command:

Terminal window
gt escalate \
--severity=<low|medium|high|critical> \
--subject="Plugin FAILED: rebuild-gt" \
--body="Build failed: make returned exit code 2" \
--source="plugin:rebuild-gt"

The command reads town config (~/gt/config.json or similar) for routing rules:

{
"escalation": {
"routes": {
"low": ["bead"],
"medium": ["bead", "mail:mayor"],
"high": ["bead", "mail:mayor", "email:human"],
"critical": ["bead", "mail:mayor", "email:human", "sms:human"]
},
"contacts": {
"human_email": "[email protected]",
"human_sms": "+1234567890"
},
"stale_threshold": "4h"
}
}
ActionBehavior
beadCreate escalation bead with severity label
mail:mayorSend mail to mayor/
email:humanSend email via configured service
sms:humanSend SMS via configured service

Every escalation creates a bead:

type: escalation
status: open
labels:
- severity:high
- source:plugin:rebuild-gt
- acknowledged:false

A patrol step (or plugin!) checks for unacknowledged escalations:

Terminal window
bd list --type=escalation --label=acknowledged:false --older-than=4h

Stale escalations get re-escalated at higher severity.

Terminal window
gt escalate ack <bead-id>
# Sets label acknowledged:true

Expose binary staleness check:

Terminal window
gt stale # Human-readable output
gt stale --json # Machine-readable
gt stale --quiet # Exit code only (0=stale, 1=fresh)

Formalized plugin dispatch to dogs:

Terminal window
gt dog dispatch --plugin <name> [--rig <rig>]

This:

  1. Finds the plugin definition
  2. Slinga a standardized work unit to an idle dog
  3. Returns immediately (non-blocking)

Unified escalation API:

Terminal window
gt escalate \
--severity=<level> \
--subject="..." \
--body="..." \
[--source="..."]
gt escalate ack <bead-id>
gt escalate list [--severity=...] [--stale]

Plugin management:

Terminal window
gt plugin list # List all plugins
gt plugin show <name> # Show plugin details
gt plugin run <name> [--force] # Manual trigger
gt plugin digest [--yesterday] # Squash wisps to digest
gt plugin history <name> # Show execution history

  1. gt stale command - Expose CheckStaleBinary() via CLI
  2. Plugin format spec - Finalize TOML schema
  3. Plugin scanning - Deacon scans town + rig plugin dirs
  1. gt dog dispatch --plugin - Formalized dog dispatch
  2. Plugin execution in dogs - Dog reads plugin.md, executes
  3. Wisp creation - Record results on ledger
  1. Gate evaluation - Cooldown via wisp query
  2. Other gate types - Cron, condition, event
  3. Plugin digest - Daily squash of plugin wisps
  1. gt escalate command - Unified escalation API
  2. Escalation routing - Config-driven multi-channel
  3. Stale escalation patrol - Check unacknowledged
  1. rebuild-gt plugin - The actual gastown plugin
  2. Documentation - So Beads/Wyvern can create theirs

+++
name = "rebuild-gt"
description = "Rebuild stale gt binary from gastown source"
version = 1
[gate]
type = "cooldown"
duration = "1h"
[tracking]
labels = ["plugin:rebuild-gt", "rig:gastown", "category:maintenance"]
digest = true
[execution]
timeout = "5m"
notify_on_failure = true
severity = "medium"
+++
# Rebuild gt Binary
Checks if the gt binary is stale (built from older commit than HEAD) and rebuilds.
## Gate Check
The Deacon evaluates this before dispatch. If gate closed, skip.
## Detection
Check binary staleness:
```bash
gt stale --json

If "stale": false, record success wisp and exit early.

Rebuild from source:

Terminal window
cd ~/gt/gastown/crew/george && make build && make install

On success:

Terminal window
bd wisp create \
--label type:plugin-run \
--label plugin:rebuild-gt \
--label rig:gastown \
--label result:success \
--body "Rebuilt gt: $OLD$NEW ($N commits)"

On failure:

Terminal window
bd wisp create \
--label type:plugin-run \
--label plugin:rebuild-gt \
--label rig:gastown \
--label result:failure \
--body "Build failed: $ERROR"
gt escalate --severity=medium \
--subject="Plugin FAILED: rebuild-gt" \
--body="$ERROR" \
--source="plugin:rebuild-gt"
---
## Open Questions
1. **Plugin discovery in multiple clones**: If gastown has crew/george, crew/max, crew/joe - which clone's plugins/ dir is canonical? Probably: scan all, dedupe by name, prefer rig-root if exists.
2. **Dog assignment**: Should specific plugins prefer specific dogs? Or any idle dog?
3. **Plugin dependencies**: Can plugins depend on other plugins? Probably not in v1.
4. **Plugin disable/enable**: How to temporarily disable a plugin without deleting it? Label on a plugin bead? `enabled = false` in frontmatter?
---
## References
- PRIMING.md - Core design principles
- mol-deacon-patrol.formula.toml - Patrol step plugin-run
- ~/gt/plugins/README.md - Current plugin stub