Track task (checkbox progress)

Use this skill when the user (or another skill) needs to track execution of a workflow: what step are we on?, mark this done, track skill A and skill B together, resume after a break.

When to activate

• User says: track this, checkpoint, what’s next, mark step N done, track [skill name], progress on …
• You are orchestrating one or more leaf skills and must persist session state as checkboxes, not by editing content/parts/process.md or phase source files.

What you do (agent behavior)

1. Resolve the engagement root
   Read skill-config.json → workspace.active_skill_workspace on the agent (or skill) root. If missing, set workspace first (see skills/workspace_skill/).

2. Pick what to track

   • Whole pipeline (abd-skill-builder style skill): phases from skill-config.json → phase_files and/or rows in content/parts/process.md.
   • Single phase: steps from ## Action Checklist or any - [ ] / - [x] lines in content/parts/phases/<slug>.md.
   • Ad-hoc / non-builder skill: build a numbered checklist from SKILL.md, README, or explicit user steps.

3. Write only to live progress files
   Normative docs stay clean. Session ticks live under:

   <active_skill_workspace>/<skill_name>/progress/
     process-checklist.md          # one line per phase (pipeline position)
     <phase-slug>-checklist.md      # steps inside that phase
   

   For builder skill packages, prefer python scripts/base/generate.py --phase <slug> to create those files when the target skill ships scripts/base/ (see Builder skill packages below). If there is no generate.py, create the same paths manually (same checkbox format).

4. Multiple skills
   One <skill_name>/progress/ tree per tracked skill. Do not merge unrelated pipelines into one file unless the user asks.

5. Each turn

   • Open the relevant progress/*.md file(s).
   • First unchecked - [ ] is “next” unless the user names a step.
   • After work completes, flip only that line to - [x].
   • Summarize: done / next / blocked.

Tracking contract (detail)

Goals: Observable - [ ] / - [x] progress; no session edits to normative content/parts/process.md or content/parts/phases/*.md; same idea for builder packages, minimal skills, and agent-only repos.

Workspace: active_skill_workspace = engagement tree (not the skill install). skill_name = skill-config.json → name if set, else the skill directory name. Live files: <workspace>/<skill_name>/progress/.

File kinds

Rules

1. Create missing progress/ files when the user agrees to start; do not overwrite existing checklists unless they ask to reset.
2. Next step = first unchecked item in scope.
3. Multi-skill: separate progress/ per skill; optional roll-up tracking-summary.md at workspace root only if requested.
4. No skill-config.json: use <workspace>/_tracking/<id>/checklist.md (id = label or folder name).

Anti-patterns: ticking boxes inside content/parts/** sources; tracking only in chat when the user wanted files across sessions.

Builder skill packages

If the tracked skill has scripts/base/generate.py and workspace_checklists.py:

1. Set workspace.active_skill_workspace for that skill (or wrapping agent).
2. Run python scripts/base/generate.py --phase <slug>.
3. First run may create process-checklist.md and <slug>-checklist.md under …/<skill_name>/progress/.

Exact creation rules: workspace_checklists.py and content/parts/library/base/checklist.md in abd-skill-builder. This skill updates those files after they exist; it does not replace generate.py. For prompt-only, no new checklists: --no-ensure-checklists.

Commands (optional helpers)

From agent root (parent of skills/):

python skills/track_task/scripts/progress_path.py
python skills/track_task/scripts/progress_path.py --skill my-other-skill

Prints …/<skill_name>/progress for the configured workspace.

abd-delivery-lead (agent checklist)

When the abd-delivery-lead agent tracks work, use abd-delivery-lead as skill_name for progress paths. Use one ad-hoc file (not process-checklist.md unless you split runs separately):

<active_skill_workspace>/abd-delivery-lead/progress/delivery-plan-checklist.md
<active_skill_workspace>/agile-delivery-plan.md

agile-delivery-plan.md (workspace root) — canonical narrative agile delivery plan (context assessment, risks, strategies, runs, checkpoints). The delivery lead and planning skill read/write this file; keep it aligned with delivery-plan-checklist.md.

Build the checklist from:

1. Orchestration — milestones that map to Steps 1–8 in agents/abd-delivery-lead/AGENT.md (workspace, planning CHECKPOINT, run summaries, final sign-off).
2. Plan — for each run from abd-delivery-planning, each stage in order: entry → bootstrap abd-team-member → exit gate → user CHECKPOINT; nest extra - [ ] lines when the run’s checkpoint policy requires in-stage checkpoints (e.g. per story).

Keep the file aligned with the current plan: when the plan is confirmed, advanced, or revised, add or adjust lines; mark completed items - [x]; do not rely on chat alone for resumable state. Stage definitions (entry/exit) stay in agents/abd-delivery-lead/stages/<stage>.md.

See also

• skills/workspace_skill/ — set active_skill_workspace before tracking.
• agents/abd-delivery-lead/AGENT.md — orchestration steps and when to populate the checklist.