execute-task

Name: execute-task
Author: anombyte93

582stars57forks

How to install execute-task?

anombyte93/prd-taskmaster/execute-task

$npx -y skills add anombyte93/prd-taskmaster --skill execute-task

Installs into the current project.

Summary

Execute the next TaskMaster task using the implementation plan with CDD verification. Picks the next ready task, matches it to the plan step, implements via a dispatched subagent, verifies subtasks with evidence, marks the task done, and loops until every task is complete. Wraps

Files · 1

View on GitHub

SKILL.md

1	# execute-task
2
3	The execution loop. Three sources converge:
4
5	- Plan (HOW) — `docs/superpowers/plans/*.md` produced by GENERATE
6	- TaskMaster (WHAT) — `.taskmaster/tasks/tasks.json` with
7	dependencies and complexity scores
8	- CDD (PROOF) — acceptance cards per task, evidence-gated
9
10	execute-task is the single skill that runs the full build from "tasks are
11	ready" to SHIP_CHECK_OK. It is autonomous — no AskUserQuestion inside the
12	loop. Any gap that would require user input is surfaced through the recon
13	escalation ladder (step 11) or the inbox (steps 4 and 8), never via a modal
14	prompt.
15
16	## Entry
17
18	This skill is invoked either:
19
20	1. Directly by the user once HANDOFF has completed and a task-execution
21	mode (A/B/C) has been dispatched, or
22	2. By the `prd-taskmaster` orchestrator when `current_phase` is `EXECUTE`.
23
24	On entry, confirm that:
25
26	- `.atlas-ai/state/pipeline.json` exists and records `phase: EXECUTE`
27	- `.taskmaster/tasks/tasks.json` exists with at least one ready task
28	- `.atlas-ai/customizations/system-prompt-template.md` is present (may be
29	empty — absence is a setup bug, empty is fine)
30
31	If any of the above are missing, report the gap and halt. Do NOT attempt to
32	bootstrap the missing artifact from inside this loop — that is the
33	orchestrator's job.
34
35	## Cycle (per iteration)
36
37	Each pass through this cycle moves exactly one TaskMaster task from `pending`
38	to `done`. Do the 13 steps in order. Do not skip.
39
40	> Task-start SHA — at the very beginning of each iteration (before step 2),
41	> capture the current git HEAD:
42	>
43	> ```bash
44	> task_start_sha=$(git rev-parse HEAD)
45	> ```
46	>
47	> Record `$task_start_sha` in the execute-log row for this iteration. It is the
48	> oracle of truth for every reachability sweep in step 9b below: "what modules
49	> did THIS task add?" is `diff $task_start_sha..HEAD`. The oracle flow already
50	> issues per-task start commits; this surfaces the same value in the loop prose.
51
52	1. Heartbeat check: verify the execute-task heartbeat timer is running.
53	If missing, register one via `CronCreate("execute-task-heartbeat", "* * * * *", "echo heartbeat")`.
54	Abort the iteration if the timer cannot be created — a missing heartbeat
55	means a missing stuck-session detector, and that is load-bearing.
56
57	2. Inbox reconciliation: read `.atlas-ai/state/pipeline.json`,
58	`.taskmaster/tasks/tasks.json`, and the current TodoWrite list.
59	Diff them. If the three are stale by more than 5 tasks (i.e. TodoWrite
60	says 10 done but tasks.json says 3 done), report the diff and halt — do
61	not paper over bookkeeping drift by silently reconciling.
62
63	3. Pick next task: run backend op `next` with the plugin's project-root
64	pointer. Use exactly this invocation:
65
66	```bash
67	python3 script.py next-task
68	```
69
70	Parse the JSON result.
71	- If no ready tasks and all tasks are `done`, run `.atlas-ai/ship-check.py`,
72	emit SHIP_CHECK_OK on success, exit the loop.
73	- If no ready tasks but pending tasks exist, the dependency graph is
74	deadlocked — report and halt.
75
76	4. Load plan step: search for the matching task ID in this priority
77	order, halting only after all three fail:
78
79	1. `docs/superpowers/plans/*.md` (the superpowers GENERATE default output)
80	2. `.taskmaster/docs/plan.md` (the prd-taskmaster HANDOFF default output,
81	whose path is also recorded in
82	`pipeline.json:phase_evidence.HANDOFF.plan_file_path`)
83	3. Any custom path declared in
84	`pipeline.json:phase_evidence.HANDOFF.plan_file_path` (in case
85	a future handoff variant writes elsewhere)
86
87	If none of the three contains the matching task ID, the task was
88	invented downstream of the plan — mark the task `blocked`, inbox the
89	parent orchestrator with `message_type="blocker"`, and continue to the
90	next iteration.
91
92	(Codified 2026-06-04 — yesterday's ai-human-tasker run had its plan at
93	`.taskmaster/docs/plan.md` only, while this step previously read
94	`docs/superpowers/plans/*.md` exclusively. The controller silently
95	improvised; a cold-start successor would have hit the `blocked` path on
96	every task.)
97
98	5. Generate CDD card: convert the task's `subtasks` field into a
99	`testing_plan`. Each subtask becomes a verifiable check with a concrete
100	evidence

People who liked this also viewed

6 picks

Type

execute-task

582stars57forks

View on GitHub

How to install execute-task?

anombyte93/prd-taskmaster/execute-task

$npx -y skills add anombyte93/prd-taskmaster --skill execute-task

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# execute-task
2
3	The execution loop. Three sources converge:
4
5	- Plan (HOW) — `docs/superpowers/plans/*.md` produced by GENERATE
6	- TaskMaster (WHAT) — `.taskmaster/tasks/tasks.json` with
7	dependencies and complexity scores
8	- CDD (PROOF) — acceptance cards per task, evidence-gated
9
10	execute-task is the single skill that runs the full build from "tasks are
11	ready" to SHIP_CHECK_OK. It is autonomous — no AskUserQuestion inside the
12	loop. Any gap that would require user input is surfaced through the recon
13	escalation ladder (step 11) or the inbox (steps 4 and 8), never via a modal
14	prompt.
15
16	## Entry
17
18	This skill is invoked either:
19
20	1. Directly by the user once HANDOFF has completed and a task-execution
21	mode (A/B/C) has been dispatched, or
22	2. By the `prd-taskmaster` orchestrator when `current_phase` is `EXECUTE`.
23
24	On entry, confirm that:
25
26	- `.atlas-ai/state/pipeline.json` exists and records `phase: EXECUTE`
27	- `.taskmaster/tasks/tasks.json` exists with at least one ready task
28	- `.atlas-ai/customizations/system-prompt-template.md` is present (may be
29	empty — absence is a setup bug, empty is fine)
30
31	If any of the above are missing, report the gap and halt. Do NOT attempt to
32	bootstrap the missing artifact from inside this loop — that is the
33	orchestrator's job.
34
35	## Cycle (per iteration)
36
37	Each pass through this cycle moves exactly one TaskMaster task from `pending`
38	to `done`. Do the 13 steps in order. Do not skip.
39
40	> Task-start SHA — at the very beginning of each iteration (before step 2),
41	> capture the current git HEAD:
42	>
43	> ```bash
44	> task_start_sha=$(git rev-parse HEAD)
45	> ```
46	>
47	> Record `$task_start_sha` in the execute-log row for this iteration. It is the
48	> oracle of truth for every reachability sweep in step 9b below: "what modules
49	> did THIS task add?" is `diff $task_start_sha..HEAD`. The oracle flow already
50	> issues per-task start commits; this surfaces the same value in the loop prose.
51
52	1. Heartbeat check: verify the execute-task heartbeat timer is running.
53	If missing, register one via `CronCreate("execute-task-heartbeat", "* * * * *", "echo heartbeat")`.
54	Abort the iteration if the timer cannot be created — a missing heartbeat
55	means a missing stuck-session detector, and that is load-bearing.
56
57	2. Inbox reconciliation: read `.atlas-ai/state/pipeline.json`,
58	`.taskmaster/tasks/tasks.json`, and the current TodoWrite list.
59	Diff them. If the three are stale by more than 5 tasks (i.e. TodoWrite
60	says 10 done but tasks.json says 3 done), report the diff and halt — do
61	not paper over bookkeeping drift by silently reconciling.
62
63	3. Pick next task: run backend op `next` with the plugin's project-root
64	pointer. Use exactly this invocation:
65
66	```bash
67	python3 script.py next-task
68	```
69
70	Parse the JSON result.
71	- If no ready tasks and all tasks are `done`, run `.atlas-ai/ship-check.py`,
72	emit SHIP_CHECK_OK on success, exit the loop.
73	- If no ready tasks but pending tasks exist, the dependency graph is
74	deadlocked — report and halt.
75
76	4. Load plan step: search for the matching task ID in this priority
77	order, halting only after all three fail:
78
79	1. `docs/superpowers/plans/*.md` (the superpowers GENERATE default output)
80	2. `.taskmaster/docs/plan.md` (the prd-taskmaster HANDOFF default output,
81	whose path is also recorded in
82	`pipeline.json:phase_evidence.HANDOFF.plan_file_path`)
83	3. Any custom path declared in
84	`pipeline.json:phase_evidence.HANDOFF.plan_file_path` (in case
85	a future handoff variant writes elsewhere)
86
87	If none of the three contains the matching task ID, the task was
88	invented downstream of the plan — mark the task `blocked`, inbox the
89	parent orchestrator with `message_type="blocker"`, and continue to the
90	next iteration.
91
92	(Codified 2026-06-04 — yesterday's ai-human-tasker run had its plan at
93	`.taskmaster/docs/plan.md` only, while this step previously read
94	`docs/superpowers/plans/*.md` exclusively. The controller silently
95	improvised; a cold-start successor would have hit the `blocked` path on
96	every task.)
97
98	5. Generate CDD card: convert the task's `subtasks` field into a
99	`testing_plan`. Each subtask becomes a verifiable check with a concrete
100	evidence

People who liked this also viewed

6 picks

Type