一键导入
dp-campaign-control
Operate SPEC-80-style campaign work in dp-codex-aware repositories; trigger on campaign, next goal, recover, block, verify, handoff.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Operate SPEC-80-style campaign work in dp-codex-aware repositories; trigger on campaign, next goal, recover, block, verify, handoff.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
Adopt or migrate old dp projects; trigger on upgrade, legacy dp, migration.
Orient in a dp-codex-aware repo; trigger on first command, opened repo, bootstrap, where to start.
Repair evidence failures; trigger on failed check, stale run, missing validator.
Manage goal state; trigger on start, block, release, verify, lease.
Audit hook failures; trigger on hook failed, Codex hook, git hook.
Respect and audit instructions; trigger on AGENTS.md, override, local law.
| name | dp-campaign-control |
| description | Operate SPEC-80-style campaign work in dp-codex-aware repositories; trigger on campaign, next goal, recover, block, verify, handoff. |
Respect AGENTS.md and any nested AGENTS.md before using this skill.
Use dp as the campaign control plane. Codex may reason and edit, but dp owns state, evidence, blocking gates, and recovery artifacts.
.beads/, dp-policy.json, or repo instructions.dp doctor --json.dp task claim --json or claim the known issue.dp campaign recover docs/campaigns/<campaign>.json --json.dp campaign run docs/campaigns/<campaign>.json --driver codex --supervised --managed --json.The managed campaign run claims at most one goal, emits the handoff package, and stops. It does not spawn Codex, execute evidence, or mark work complete.
read_first files before editing.dp goal start docs/goals/<goal>.json --agent codex --json.make check.dp evidence run docs/evidence/<evidence>.json --output docs/evidence-runs/<run>.json --json when the goal has an evidence plan.dp verify --goal docs/goals/<goal>.json --evidence docs/evidence-runs/<run>.json --json before claiming completion.dp campaign sync-beads docs/campaigns/<campaign>.json --write --json when campaign state needs to reconcile back to Beads.Block through dp when a required spec, decision, validator, or scope change is missing:
dp goal block docs/goals/<goal>.json --reason needs_decision --write-artifact --json
Use the GoalContract blocker routes: needs_specification, needs_decision, needs_validator,
unsafe_scope, or budget_exhausted.
Release the goal instead of leaving an invisible claim when stopping without completion:
dp goal release docs/goals/<goal>.json --reason "context reset" --json
Before ending the session:
make check.dp doctor --json.bd --readonly status --json.git status reports the branch is up to date with origin.Do not mark completion from agent narration. Do not execute raw shell from generated JSON. Do not call an LLM from hooks, validators, evidence assertions, or CI. Do not spawn Codex from dp in this workflow. Do not replace Beads with campaign files; synchronize through explicit dp or bd commands.