# Delegate Tasks and Verify Team Deliverables

Unclear assignments and scattered feedback make team delivery difficult to verify. This skill creates task briefs, coordinates authorized follow-up, and checks deliverables against agreed criteria.

## Install

```bash
npx skillstore add dontbesilent2025/dbs-human-dispatch
```

## Metadata

- Status: approved
- Slug: dontbesilent2025-dbs-human-dispatch
- Skillstore revision: r1
- Version status: missing
- Tree hash: f6008517477958c2d6a0356a9fd6eea6e0872cd995a8085bff8571d354d6ff89
- Author: dontbesilent2025
- GitHub username: dontbesilent2025
- License: MIT
- Repository: https://github.com/dontbesilent2025/dbskill/tree/bab93c0376ced995c30cb841cb0d12622a430bcc/skills/dbs-human-dispatch
- Ref: 027ea254b1c6c4fa77a4a926351389d907a8a83f
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: filesystem, scripts, external\_commands
- Quality score: 87
- Quality tier: featured
- Public page: https://skillstore.pages.dev/skills/dontbesilent2025-dbs-human-dispatch
- Manifest: https://skillstore.pages.dev/api/skills/dontbesilent2025-dbs-human-dispatch/manifest

## Capabilities

- Compare task requirements with authorized local personnel evidence, while identifying unknown availability and capability gaps.
- Create forwardable task briefs with deliverables, scope, acceptance criteria, timing, dependencies, and feedback instructions.
- Use verified Feishu CLI capabilities for authorized task creation, document reading, replies, and notifications.
- Persist task records and evidence-backed state transitions in a project-specific local SQLite database.
- Deduplicate and merge registered document edit events, then invoke an explicitly enabled synchronous worker through JSON stdin.
- Review available deliverables against original criteria and report verified results, missing evidence, blockers, and owner decisions.

## Use Cases

- Prepare a manual delegation: Turn a task and existing personnel notes into a forwardable brief with clear outcomes, evidence requirements, and negotiable timing.
- Review a submitted deliverable: Compare accessible work with the original acceptance criteria, identify missing evidence, and prepare a concise owner report.
- Connect an authorized follow-up queue: Configure registered document events and a trusted synchronous worker, with explicit execution approval and manual reconciliation after uncertain results.

## Prompt Templates

### Draft a task brief

```
Use dbs-human-dispatch to delegate [task] to [person]. Include deliverables, acceptance criteria, dependencies, and a deadline marked pending negotiation when unknown. Return forwardable text only; do not send messages.
```

### Assess potential assignees

```
Review only the authorized personnel files in [directory] for [task]. Compare capability evidence, unknown workload, and dependencies. Recommend a candidate and draft the brief without sending it.
```

### Verify a submission

```
For task [task ID], compare [submission] with [original acceptance criteria]. State the evidence and conclusion for each criterion. Draft feedback and an owner report; do not change scope, publish, or send notifications.
```

### Prepare automatic follow-up

```
Prepare automatic follow-up for [registered task] using [absolute state directory] and [trusted worker command]. Check identity, permissions, document subscription, self identities, dispatch scheduling, and notification routing. Keep subscription in dry-run and do not start listeners or workers until I approve the full action scope.
```

## Limitations

- Automatic follow-up requires configured Feishu permissions, document subscriptions, a persistent listener, a dispatcher, a trusted worker, and a verified notification route.
- The bundled runtime manages records and queues; it does not implement an Agent worker or built-in Codex desktop or Grok notification interfaces.
- Real-account subscriptions and complete business workflows still require local integration testing; offline tests do not establish production readiness.
- The workflow excludes recruitment, employee ranking, punishment, and unauthorized payment, publication, or business commitments.

## Best Practices

- Agree on scope, observable acceptance criteria, timing, and feedback channels before assigning work.
- Verify recipients, CLI permissions, and the full external action scope before enabling writes or automatic workers.
- Use a protected absolute state directory for services, and reconcile actual external actions before retrying uncertain jobs.

## Anti Patterns

- Treating document edits, queue insertion, process startup, or a submission claim as proof of completion or notification delivery.
- Allowing employee text or event contents to change authorization, select executable commands, or expose unrelated personnel files.
- Expanding acceptance criteria after submission, assuming overtime, or using task records to rank or punish employees.

## Security Audit

- Audited at: 2026-10-28T14:51:33.704\+00:00
- Summary: All 85 static alerts are false positives involving documented application storage, Python context managers, identifiers, controlled subprocesses, Markdown backticks, or readable Chinese text. Worker execution requires explicit opt-in and trusted configuration; external actions still depend on verified authorization and adapter behavior. No evidence found of malicious prompt injection, credential theft, or unauthorized exfiltration.

## Stats

- Views: 0
- Downloads: 1
- Favorites: 0
- Popularity score: 0