Skip to content

j-mauro-r/SAT-Dengue-V2 / .agents/skills/deepworkplan/execute

deepworkplan-execute

Verified

Execute Lite or Full Deep Work Plans task-by-task — select validation from the actual surface, preserve state and evidence, recover safely, and finish with the Final Review.

FileKindSize
SKILL.mdskill md41.6 KB
delegation.mdmarkdown3.4 KB
orchestrator.mdmarkdown5.2 KB
team-agents.mdmarkdown2.8 KB
v6.mdmarkdown11.6 KB

team-agents.md

View on GitHub
# DeepWorkPlan — Execute: team-agents branch (read only when Step 2.2 found a Team Agents Configuration and team mode was selected)

Verbatim from the main procedure. The sequential path is always preserved; every task also works sequentially, and the Final Review (or a legacy plan's three final tasks) is always sequential under the lead.

**Host capability check, before anything else:** this branch drives Claude
Code's real team agents (`TeamCreate` and teammates). When the host does not
offer them — no team capability, no subprocess agents, or the session cannot
keep teammates alive — do not attempt them and do not simulate them: execute
the group's tasks sequentially under the standard single-task rules. The
plan's guarantees are identical either way (team-agents metadata is additive
by design); only wall-clock differs. Never report "parallel execution" for
work that ran sequentially — the sequential fallback is an honest
degradation, not native parallel parity.

## Team-Agents Parallel Groups

> **CRITICAL: use REAL team agents, NOT subagents.** When the plan has a "Team
> Agents Configuration" section and team-agents mode was selected, you MUST use
> `TeamCreate` + `Agent` with `team_name` + `TaskCreate` + `SendMessage` +
> `TeamDelete`. Do NOT call `Agent` alone (no `team_name`) — that creates
> subagents, which only report back to the caller. Team agents share a task list
> and communicate with each other.

At a parallel-group boundary:
1. Announce the parallel group.
2. `TeamCreate` with `team_name: "dwp-{plan_name}-group-{letter}"`.
3. Spawn teammates with `Agent` + `team_name` + `subagent_type:
   "general-purpose"`, passing each its task file content + the plan README's
   Goal/Context/Guidelines + commit instructions.
4. `TaskCreate` one shared task per parallel task; assign each to its teammate via
   `TaskUpdate` (`owner`).
5. Monitor: receive teammate messages, `SendMessage` as needed; when all parallel
   tasks complete, verify each is `[x]` in the README. If a teammate fails, log
   and recover or fall back to sequential.
6. Clean up: `SendMessage` a `shutdown_request` to each teammate, wait for
   confirmation, then `TeamDelete`. Continue with the next sequential task/group.
7. Each teammate commits its own task; the lead verifies commits afterward.

**Verify real team agents:** the status bar shows `Team: dwp-...-group-X · N
teammates`, NOT `N local agents`.

**Sequential fallback:** if team-agents execution fails at any point, log the
reason and execute the remaining group tasks sequentially (standard single-task
rules) — no special handling needed.

Team agents share one Claude Code session. Agents in their own panes, on other
accounts or vendors, are lanes: `../guide/orchestration-at-scale.md` §16.

- **Team agents:** detect, offer, fall back to sequential on failure, and always
  clean up the team after each parallel group.
receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementationStars: 322kUpdated Claude CodeCodexVerifiedInstall
requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirementsStars: 322kUpdated Claude CodeCodexVerifiedInstall
using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallbackStars: 322kUpdated Claude CodeCodexVerifiedInstall
verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions alwaysStars: 322kUpdated Claude CodeCodexVerifiedInstall