Human

Manages tasks that require human intervention after all automation attempts have failed.

hivetechs-collective 1 updated 1mo ago
Claude CodeGeneric
View source ↗
# Human Action Queue Management

Manages tasks that require human intervention after all automation attempts have failed.

## Usage

/human # Show current queue with stats /human complete # Mark an action as complete /human add # Manually add an action item /human blocked # Show all tasks blocked by human actions


## Command Actions

### Default: Show Queue

Display the current @human_actions.md with live statistics:
- Pending actions count
- Total blocked tasks
- Priority breakdown (Critical/High/Medium/Low)

```bash
pnpm run human:status

Complete Action

Mark a human action as completed:

  1. Read @human_actions.md to show pending actions
  2. Ask user which action was completed
  3. Change - [ ] to - [x] for that action
  4. Run sync to update blocking relationships
  5. Report how many tasks are now unblocked
pnpm run human:complete HA-XXX

Add Action

Manually add a new human action item:

  1. Ask for action title
  2. Ask for instructions
  3. Ask what tasks it blocks (if known)
  4. Generate unique ID (HA-XXX)
  5. Add to @human_actions.md with proper format

Show Blocked

Display all tasks in fix_plan.md that are blocked by human actions:

# Check fix_plan.md for [!] blocked items
grep "^\- \[!\]" fix_plan.md

Integration with Ralph

The human action queue integrates with the development pipeline:

  1. Detection: Ralph detects when a task requires human intervention
  2. Pre-check: Ralph MUST try automation first (CLI, Playwright, agents)
  3. Escalation: Only after automation fails, add to @human_actions.md
  4. Blocking: Related tasks in fix_plan.md are marked blocked
  5. Completion: Human marks done, blocked tasks become actionable
  6. Resume: Ralph automatically picks up newly unblocked tasks

Automation First (MANDATORY)

Before adding ANY human action, verify these have been tried:

  1. CLI Tools: wrangler, gh, stripe, neonctl
  2. Environment: Check .env files for existing secrets
  3. Playwright: Automate browser dashboards
  4. Agents: cloudflare-expert, devops-automation-expert, etc.
  5. Documentation: /docs/guides/setup/ for alternatives

See /docs/operations/AUTOMATION_CAPABILITIES.md for full reference.

File Locations

  • Queue: @human_actions.md (repo root)
  • Script: scripts/governance/human-actions.ts
  • Docs: docs/operations/AUTOMATION_CAPABILITIES.md

Examples

Viewing the Queue

/human

Output:
# Human Action Queue
**Pending Actions:** 3 | **Total Blocked Tasks:** 47

## 🔴 CRITICAL (Blocking 10+ tasks)
### [HA-001] Enable R2 Storage in Cloudflare Dashboard
...

Completing an Action

/human complete

> Which action did you complete?
> HA-001

✅ Marked HA-001 as complete
   Unblocked 23 tasks in fix_plan.md
   Next highest priority: HA-002 (Blocking 12 tasks)

Adding an Action

/human add

> Title: Configure Apple Developer Certificate
> Instructions:
>   1. Log into Apple Developer Portal
>   2. Create new Developer ID certificate
>   3. Download and install in Keychain
> Blocked tasks: macOS signing, notarization

✅ Created HA-003: Configure Apple Developer Certificate
   Priority: P1 (Blocking 8 tasks)

Maintain Human?

Let people know it's listed here — add the badge (live metrics, light/dark aware) or a plain link to your README or docs.

[Human on getagentictools](https://getagentictools.com/loops/hivetechs-collective-human-action-queue-management?ref=badge)