Direct execution — Forge Sparks and Charge candidates

Related: Forge — major processes & flow maps · Epic execution profile (Epic execution profile) · Daily operations (forge/charge.md) · Forge & planning — naming reference (Forge Spark vs Product Spark) · Markdown-canonical workspace policy (optional repo profile) · Documentation structure — proposal

---

Canonical paths (stock blueprint)

Team override: If docs/PROJECT.md documents forge/current/CHARGE.md and forge/current/ITERATION.md (or similar), use those paths instead and state that in the change plan. Do not introduce forge/current/ without recording it in docs/PROJECT.md.

---

How to use

1. Run the classifier prompt; confirm direct execution (or equivalent) in output A.
2. Copy from “Act as Forge execution planner…” through “F. Risks…” below.
3. Paste the request into the fence.
4. Prefer forge/charge.md unless the repo declares an alias.
5. Never write project history under blueprints/.

---

Prompt body (copy from here)

Act as Forge execution planner for this workspace.

Convert this direct executable request into committed work and decide Charge membership.

Under the Epic execution profile: execute inside the Charged Epic — apply OpenSpec acceptance, decompose runs at L1–L2, do not mint Forge Sparks or …T{n} rows on Charge. L1 trivial skip: obvious one-liners may bypass Epic ceremony (no OpenSpec folder, no Charge row). Core Forge (default): convert to Forge Sparks as below.

Request:

<<<PASTE REQUEST HERE>>>

Workspace rules:

• Markdown-only canonical artifacts (.md); no CSV / spreadsheet SoT.
• Dual profile: Core Forge keeps Forge Spark → Charge. Under the Epic execution profile, Charge = Epics with OpenSpec acceptance — Story/Task/Spark are agent scratch inside the Epic, not Charge items.
• Forge Spark (core only) = smallest delivery unit (~1–4 h), often WBS task id M{n}E{n}S{n}T{n}; not a Product Spark (release slice).
• Charge lists today's committed work — forge/charge.md: Sparks (core) or Epics (profile); see team alias in docs/PROJECT.md if documented.

Tasks (core Forge — Forge Sparks):

1. Decide whether the request is one Forge Spark or several (split if multiple PR-sized outcomes, different repos, or non-overlapping acceptance).

2. Identify prerequisites, sequencing, and hidden decomposition (unknowns, discovery work — if the “hidden” work is learning-only, flag as discipline exploration spike per Discipline exploration spike — lifecycle and anchors, not a Forge Spark).

3. Search for existing same/similar Spark-level items in:

4. docs/requirements/WBS.md
5. Linked task / Spark Markdown files (e.g. docs/requirements/milestones/.../tasks/*-task.md or team sparks/*.md)
6. Charge file: forge/charge.md or team alias forge/current/CHARGE.md if documented
7. Iteration summary: forge/current/ITERATION.md if present, else docs/requirements/WBS.md / docs/ROADMAP.md / relevant forge/releases/*.md

8. docs/requirements/TRACEABILITY.md

9. For each candidate Forge Spark, choose exactly one canonicalization action A–G from Markdown-canonical workspace policy (optional repo profile).

10. Propose or update Spark records (front matter and/or body) with:

11. id
12. title
13. parent — Ingot / story id (M{n}E{n}S{n}) and, if known, Product Spark / milestone link
14. repos / module affected
15. Acceptance criteria (testable)
16. dependencies (Spark ids or external)
17. Owner discipline / suggested Versona (e.g. SE, UX, DevOps)
18. estimation placeholder (TBD or team format)

19. canonicalization (action A–G + note)

20. Update or propose Markdown updates to:

21. docs/requirements/WBS.md
22. Spark / task files under docs/requirements/...
23. docs/requirements/TRACEABILITY.md
24. Iteration doc if the repo maintains forge/current/ITERATION.md; otherwise the WBS / ROADMAP section that names the active Forge iteration

25. forge/charge.md (or documented Charge alias)

26. Recommend which Sparks belong on the next Charge (today / next session) vs backlog — respect WIP and dependencies.

27. Flag any item too large for one Forge Spark (suggest split into multiple …T{n} or refinement into an Ingot first).

Tasks (Epic execution profile — inside Charged Epic):

1. Confirm target Epic on Charge (Active Epics) or classify as L1 trivial skip (no Epic row).
2. Load OpenSpec change — observable SHALLs + scenarios; tasks.md is non-binding.
3. Plan runs inside the Epic (L1–L2); do not mint …T{n} on Charge or expand WBS Task process for the Epic.
4. Escalate L4+ (cross-repo L4.2, ADR, security) — stop and ask before proceeding.
5. Update traceability at Epic scope; recommend Charge status transitions only for the Epic row.

Execution mode:

• High confidence — apply direct Markdown edits.
• Lower confidence — output exact file paths and full Markdown fragments for review.

Return (structured output):

Profile detected: core Forge · Epic execution profile · L1 trivial skip

A. Work list — Core: Spark table (id, title, phase prefix, status, parent). Profile: Epic id, OpenSpec change, run plan inside Epic (no Charge Spark rows).
B. Sequence and dependencies — ordered graph or numbered list.
C. Charge recommendation — Core: which Spark ids to add/remove on today's Charge. Profile: Epic status/actor updates only.
D. Markdown file changes — path + create/update + one-line purpose.
E. Canonicalization decisions — per candidate: A–G, similarity state, ledger pointer (TRACEABILITY / IMPORT-LEDGER).
F. Risks and hidden decomposition — bullets: unknowns, spikes, oversized work, cross-repo coupling, L4+ escalation.

---

Maintainer notes

• Scripts (forge-charge.sh) expect forge/charge.md — if using forge/current/CHARGE.md, either symlink, wrap the script, or manually mirror; document in docs/PROJECT.md.
• After Charge update, daily sync alignment is per Forge — ceremonies & events (prescriptive) (C3).
• If no parent Ingot exists, either link Spark to the nearest story in WBS or trigger Meta-request decomposition — roadmap / WBS / Forge artifacts to create the Ingot row first.

Related pages

• Forge workspace
• Assay Gate — {Product Spark name}
• Charge — YYYY-MM-DD
• Directive Change Proposal — DCP-<YYYYMMDD>…
• Discipline exploration spike — close record