Skip to content

Commit 3d20d26

Browse files
e0ipsoe0ipso
committed
feat: transition to AI TM 2
1 parent f265bb4 commit 3d20d26

7 files changed

Lines changed: 1322 additions & 0 deletions

File tree

‎.agents/skills/task-create-plan/SKILL.md‎

Lines changed: 120 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,120 @@
1+
---
2+
name: task-create-plan
3+
description: Create a new AI Task Manager plan for this repository. Use when the user asks to draft, plan, or scope a new task-manager plan — discovers the local .ai/task-manager root, runs the project's plan hooks, gathers clarifications, allocates the next plan ID, and writes a semantic HTML plan conforming to PLAN_TEMPLATE.html. Do not use for generic brainstorming or work outside the AI Task Manager.
4+
---
5+
6+
# task-create-plan
7+
8+
Drive the end-to-end creation of a new AI Task Manager plan for the user's
9+
repository. The skill is assistant-agnostic and self-contained: every script
10+
it invokes lives under this skill's `scripts/` directory and is referenced
11+
by relative path.
12+
13+
## Inputs
14+
15+
The user's request supplies the work order. Treat it as the only authoritative
16+
source of intent. Do not invent answers to clarifying questions — prompt the
17+
user instead.
18+
19+
## Operating Procedure
20+
21+
### 1. Locate the task-manager root
22+
23+
Run `scripts/find-task-manager-root.cjs` from the user's working directory.
24+
The script walks up looking for `.ai/task-manager/.init-metadata.json` and
25+
prints the absolute path of the resolved root on success.
26+
27+
If the script exits non-zero, the working directory is not inside an
28+
initialized task-manager workspace. Stop and ask the user to run the project
29+
initializer (e.g. `npx @e0ipso/ai-task-manager init`) before continuing. Do
30+
not attempt to create a plan outside of a valid root.
31+
32+
For every subsequent step, treat the path printed by this script as `<root>`.
33+
34+
### 2. Load project context
35+
36+
Read `<root>/config/TASK_MANAGER.md` for the directory structure conventions
37+
this project uses. Read `<root>/config/hooks/PRE_PLAN.md` and execute the
38+
instructions it contains before proceeding. Read
39+
`<root>/config/templates/PLAN_TEMPLATE.html` so the plan you emit conforms
40+
to its semantic structure.
41+
42+
### 3. Analyze the work order
43+
44+
Identify:
45+
46+
- Objective and end goal.
47+
- Scope and explicit boundaries.
48+
- Success criteria.
49+
- Dependencies, prerequisites, blockers.
50+
- Technical requirements and constraints.
51+
52+
### 4. Clarification loop
53+
54+
If any critical context is missing, ask the user targeted questions. Keep
55+
looping until you have no further questions. Explicitly confirm whether
56+
backwards compatibility is required. Never invent answers; never paper over
57+
a missing answer.
58+
59+
If the user declines to clarify a blocking question, stop and report the
60+
plan as needing clarification. Do not produce a partial plan.
61+
62+
### 5. Allocate the next plan ID
63+
64+
Run `scripts/get-next-plan-id.cjs` to obtain the next available plan ID.
65+
Pass `<root>` as the first argument when invoking the script from a working
66+
directory that is not inside the project, otherwise no argument is required.
67+
The script prints a single integer.
68+
69+
Compute the zero-padded form for directory naming (`{padded-id}--{slug}`)
70+
and use the unpadded integer in the plan frontmatter and the final summary.
71+
72+
### 6. Emit the plan
73+
74+
Write the plan to:
75+
76+
```
77+
<root>/plans/{padded-id}--{slug}/plan-{padded-id}--{slug}.html
78+
```
79+
80+
The output must:
81+
82+
- Conform to `<root>/config/templates/PLAN_TEMPLATE.html`, including required
83+
`<meta>` elements (at minimum `id`, `summary`, `created`) inside `<head>`.
84+
- Contain the standard sections from the template body.
85+
- Use semantic HTML, not free-form prose.
86+
- Avoid time estimates, task lists, or code samples — those belong to the
87+
later task-generation phase.
88+
89+
The `<slug>` is derived from the plan summary: lowercase, alphanumeric and
90+
hyphens only, collapsed, trimmed.
91+
92+
### 7. Run post-plan hook
93+
94+
Execute `<root>/config/hooks/POST_PLAN.md` after the plan file is written.
95+
96+
### 8. Emit the structured summary
97+
98+
Conclude with exactly this block as the final output:
99+
100+
```
101+
---
102+
103+
Plan Summary:
104+
- Plan ID: [numeric-id]
105+
- Plan File: [absolute-path-to-plan-file]
106+
```
107+
108+
The summary is consumed by downstream automation; keep the format exact.
109+
110+
## Failure Modes
111+
112+
- **No task-manager root found.** Stop, instruct the user to initialize the
113+
project. Do not write any files.
114+
- **User refuses to answer a clarifying question that blocks planning.**
115+
Report `needs-clarification` and stop. Do not produce a plan.
116+
- **Plan ID script fails.** Re-check the resolved root and re-run. If it
117+
continues to fail, surface stderr to the user and stop — do not guess an ID.
118+
- **Plan directory already exists for the allocated ID.** Re-run the
119+
next-plan-id script (a concurrent run may have advanced it) and retry once.
120+
If the conflict persists, stop and report.

‎.agents/skills/task-execute-blueprint/SKILL.md‎

Lines changed: 139 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,139 @@
1+
---
2+
name: task-execute-blueprint
3+
description: Execute an AI Task Manager plan blueprint for this repository. Use when the user asks to run, implement, or carry out a specific plan ID — discovers the local .ai/task-manager root, resolves the plan, validates or auto-generates tasks and the execution blueprint, optionally creates a feature branch, runs phases with lifecycle hooks, enforces validation gates, appends an execution summary, and archives the completed plan. Do not use for generic development work outside the AI Task Manager.
4+
---
5+
6+
# task-execute-blueprint
7+
8+
Drive the end-to-end execution of an existing AI Task Manager plan blueprint. The skill is assistant-agnostic and self-contained: every script it invokes lives under this skill's `scripts/` directory and is referenced by relative path.
9+
10+
## Critical Rules
11+
12+
1. **Never skip validation gates** — a phase is not complete until `POST_PHASE.md` succeeds.
13+
2. **Preserve dependency order** — never execute a task before all of its dependencies are completed.
14+
3. **Maximize parallelism within each phase** — run all tasks whose dependencies are satisfied simultaneously.
15+
4. **Fail safely and document everything** — halt on unrecoverable errors, and record all decisions, issues, and outcomes under "Noteworthy Events" in the execution summary.
16+
17+
## Inputs
18+
19+
The user supplies the numeric plan ID conversationally. Treat it as the only authoritative source of intent. Do not invent answers to clarifying questions — prompt the user instead.
20+
21+
## Operating Procedure
22+
23+
### 1. Locate the task-manager root
24+
25+
Run `scripts/find-task-manager-root.cjs` from the user's working directory. The script walks up looking for `.ai/task-manager/.init-metadata.json` and prints the absolute path of the resolved root on success.
26+
27+
If the script exits non-zero, the working directory is not inside an initialized task-manager workspace. Stop and ask the user to run the project initializer (e.g. `npx @e0ipso/ai-task-manager init`) before continuing. Do not attempt to execute a plan outside of a valid root.
28+
29+
For every subsequent step, treat the path printed by this script as `<root>`.
30+
31+
### 2. Resolve the plan
32+
33+
Run `scripts/validate-plan-blueprint.cjs <plan-id> planFile` to obtain the absolute path of the plan file. The same script also accepts these field names (single-field output mode) and exposes them on demand:
34+
35+
- `planDir` — absolute path of the plan directory
36+
- `taskCount` — number of existing task files in that plan's `tasks/`
37+
- `blueprintExists` — `yes` or `no`
38+
- `taskManagerRoot` — absolute path of `<root>`
39+
- `planId` — the resolved numeric plan ID
40+
41+
If the script exits non-zero, stop and ask the user to confirm the plan ID. Do not guess a different ID.
42+
43+
### 3. Validate tasks and blueprint existence
44+
45+
Inspect the `taskCount` and `blueprintExists` values returned by the validation script.
46+
47+
### 4. Auto-generate tasks and blueprint if missing
48+
49+
If `taskCount` is 0 or `blueprintExists` is `no`:
50+
51+
- Notify the user: "Tasks or execution blueprint not found. Generating tasks automatically..."
52+
- Follow the `task-generate-tasks` skill for this plan ID. Execute its operating procedure in full, including running `POST_TASK_GENERATION_ALL.md` to produce the execution blueprint.
53+
- After generation completes, re-run `scripts/validate-plan-blueprint.cjs <plan-id> planFile` (and the other fields) to refresh the resolved paths and counts.
54+
55+
If generation still leaves the plan without tasks or a blueprint, stop and report failure. Do not attempt execution without a valid blueprint.
56+
57+
### 5. Optionally create a feature branch
58+
59+
Run `scripts/create-feature-branch.cjs <plan-id>`. The script creates a branch named after the plan and prints the branch name. Continue execution regardless of whether a branch is created (some projects may skip this step).
60+
61+
### 6. Load project context and execution blueprint
62+
63+
Read these files, in order:
64+
65+
- `<root>/config/TASK_MANAGER.md` — directory conventions and project context.
66+
- The plan document at the path returned by step 2.
67+
- The plan's Execution Blueprint section — this defines the phase groupings and task dispatch order.
68+
69+
### 7. Execute phases in order
70+
71+
Use an internal task or todo tracker to monitor progress. For each phase defined in the Execution Blueprint:
72+
73+
#### 7a. Phase pre-execution
74+
Read `<root>/config/hooks/PRE_PHASE.md` and execute its instructions before starting the phase.
75+
76+
#### 7b. Task dispatch
77+
Identify all tasks scheduled for this phase whose dependencies are fully satisfied. Read `<root>/config/hooks/PRE_TASK_EXECUTION.md` and execute its instructions before starting any implementation work.
78+
79+
Deploy all selected agents simultaneously using your internal Task tool. Each agent MUST:
80+
81+
1. Read and execute `<root>/config/hooks/PRE_TASK_EXECUTION.md` before starting any implementation work.
82+
2. Execute the task according to its requirements.
83+
3. Monitor execution progress and capture outputs and artifacts.
84+
4. Update task status in real-time.
85+
86+
Maximize parallelism within each phase. Run every task that is ready at the same time.
87+
88+
#### 7c. Phase completion verification
89+
Ensure every task in the phase has status `completed`. Collect and review all task outputs. Document any issues or exceptions encountered.
90+
91+
#### 7d. Phase post-execution
92+
Read `<root>/config/hooks/POST_PHASE.md` and execute its instructions. Do not proceed to the next phase until this hook succeeds.
93+
94+
Update the phase status to `completed` in the plan's Execution Blueprint section.
95+
96+
Repeat for the next phase until all phases are complete.
97+
98+
### 8. Post-execution validation
99+
100+
Read `<root>/config/hooks/POST_EXECUTION.md` and execute its instructions. If validation fails, halt execution. The plan remains in `plans/` for debugging.
101+
102+
### 9. Append execution summary
103+
104+
Append an execution summary section to the plan document using the format described in `<root>/config/templates/EXECUTION_SUMMARY_TEMPLATE.md`. Populate:
105+
106+
- **Status**: Completed Successfully
107+
- **Completed Date**: current date
108+
- **Results**: brief summary of deliverables
109+
- **Noteworthy Events**: all decisions, issues, and outcomes encountered during execution. If none occurred, state "No significant issues encountered."
110+
- **Necessary follow-ups**: any follow-up actions or optimizations
111+
112+
### 10. Archive the plan
113+
114+
Move the completed plan directory from `<root>/plans/<plan-folder>` to `<root>/archive/<plan-folder>`.
115+
116+
Preserve the entire folder structure (including all tasks and subdirectories) to maintain referential integrity. If the move fails, log the error but do not fail the overall execution — the implementation work is complete.
117+
118+
## Failure Modes
119+
120+
- **No task-manager root found.** Stop and instruct the user to initialize the project. Do not execute any tasks.
121+
- **Plan ID does not resolve.** Stop and surface the script's stderr to the user. Do not guess a different ID.
122+
- **Missing blueprint after auto-generation.** If the `task-generate-tasks` skill fails to produce tasks or a blueprint, stop and report failure. Do not attempt execution without a blueprint.
123+
- **Hook failure.** If `PRE_PHASE.md`, `POST_PHASE.md`, or `POST_EXECUTION.md` fails, halt execution. The plan remains in `plans/` for debugging and potential re-execution.
124+
- **Execution errors.** If a task fails, read `<root>/config/hooks/POST_ERROR_DETECTION.md`, document the error in Noteworthy Events, halt the phase, and request user direction before continuing.
125+
126+
## Execution Summary
127+
128+
Conclude with exactly this block as the final output:
129+
130+
```
131+
---
132+
Execution Summary:
133+
- Plan ID: [numeric-id]
134+
- Status: Archived
135+
- Location: [absolute path to archive directory]
136+
---
137+
```
138+
139+
The summary is consumed by downstream automation; keep the format exact.

0 commit comments

Comments
 (0)