aibdd-form-activity

name	aibdd-form-activity
description	將上游 `/aibdd-flows-specify` 產出的 Activity modeling elements 逐一翻譯成 `.activity` DSL，寫入檔案並執行語法驗證。
metadata	{"user-invocable":false,"source":"project-level dogfooding"}

aibdd-form-activity

Formulation skill。只負責把 /aibdd-flows-specify 的 02-activity-analyze 已完成的 Activity modeling elements 轉成 .activity 語法；不重新判斷 actor 合法性、action 顆粒度、reuse、operation partition、或需求語意。

§1 REFERENCES

ID	Path	Phase scope	Purpose
R1	aibdd-core::ssot/spec-package-paths.md	global	kickoff boundary-aware path SSOT
R2	references/role-and-contract.md	Phase 1	caller payload schema + role boundary
R3	references/format-reference.md	Phase 2	Activity element → .activity syntax mapping
R4	scripts/decoder.py	Phase 4	.activity DSL decoder（同時為 SKILL.md Phase 4 syntax validator 與 BDD subject；spec SSOT = scripts/tests/activity-decode.feature + scripts/tests/activity-benchmark.feature）

§2 SOP

Phase 1 — ASSERT intake｜驗證 caller payload

produces: $$payload, $$target_path, $$format, $$activity

1. $contract = READ references/role-and-contract.md
2. $$payload = READ caller payload
3. $$target_path = PARSE $$payload.target_path
4. $$format = PARSE $$payload.format
5. $reasoning = PARSE $$payload.reasoning
6. $$activity = PARSE $reasoning.activity_analysis.activity
7. $graph_gaps = PARSE $reasoning.activity_analysis.graph_gaps
8. $exit_status = PARSE $reasoning.activity_analysis.exit_status
9. $payload_ok = JUDGE $$payload against $contract
10. ASSERT $payload_ok
11. ASSERT $$target_path non-empty
12. ASSERT $$format == ".activity"
13. ASSERT $$activity.name non-empty
14. ASSERT $$activity.initial exists
15. ASSERT count($$activity.finals) >= 1
16. ASSERT every element in $$activity.nodes is one of Action | Decision | Fork | Merge | Join
17. ASSERT $exit_status == "complete"
18. ASSERT count($graph_gaps) == 0

Phase 2 — RENDER element by element｜轉成 Activity DSL

produces: $$activity_doc

1. $syntax_map = READ references/format-reference.md
2. $header = RENDER [ACTIVITY] ${$$activity.name}
3. $actors = RENDER one [ACTOR] line per $$activity.actors[]
4. $initial = RENDER [INITIAL]
5. $node_lines = DERIVE empty line list
6. LOOP per $node in $$activity.nodes 6.1 BRANCH $node.type ui_step | command | query: 6.1.1 $line = RENDER Action as [STEP:${$node.display_id}] @${actor.name} ${$node.name} {${$node.binds_feature}} decision: 6.1.2 $line = RENDER Decision as [DECISION:${$node.display_id}] ${$node.condition} 6.1.3 $path_lines = RENDER every decision_path as [BRANCH:${$node.display_id}:${path.guard}] fork: 6.1.4 $line = RENDER Fork as [FORK:${$node.display_id}] 6.1.5 $path_lines = RENDER every fork_path as [PARALLEL:${$node.display_id}] merge: 6.1.6 $line = RENDER Merge as [MERGE:${$node.display_id}] join: 6.1.7 $line = RENDER Join as [JOIN:${$node.display_id}] 6.2 $node_lines = DERIVE append $line and $path_lines END LOOP
7. $finals = RENDER one [FINAL] line per $$activity.finals[]
8. $$activity_doc = RENDER .activity document from $header, $actors, $initial, $node_lines, and $finals
9. ASSERT $$activity_doc contains [ACTIVITY], [INITIAL], and [FINAL]

Phase 3 — WRITE artifact｜寫入 Activity 檔

produces: $$written_paths

1. $target_exists = MATCH path_exists($$target_path)
2. IF $target_exists ∧ $$payload.mode != "overwrite": 2.1 $msg = DRAFT "path 衝突：target_path 已存在，caller 需指定 mode=overwrite 或改 target_path" 2.2 RETURN $msg
3. WRITE $$target_path ← $$activity_doc
4. $$written_paths = DERIVE [$$target_path]
5. ASSERT $$target_path in $$written_paths

Phase 4 — VALIDATE syntax｜語法驗證

produces: $$validation_report

1. $validator = COMPUTE scripts/decoder.py
2. $validation_out = TRIGGER python3 ${$validator} ${$$target_path}
3. $$validation_report = PARSE $validation_out, schema=activity-syntax-validation-report
4. ASSERT $$validation_report.ok == true

Phase 5 — RETURN report｜回傳 caller 訊號

produces: $$report

1. $$report = DRAFT report JSON ← {status: "completed", target_path: $$target_path, written_paths: $$written_paths, syntax_valid: $$validation_report.ok, validation_report: $$validation_report}
2. RETURN $$report

§3 FAILURE & FALLBACK

Phase 1 fail handling

• IF caller payload 完全空: RETURN "empty payload — caller misuse"
• IF target_path 未指定: RETURN "target_path missing"
• IF format != ".activity": RETURN "format unsupported"
• IF reasoning.activity_analysis.activity 缺失: RETURN "activity modeling elements missing"

Phase 2 fail handling

• IF Activity element 無法映射到 .activity tag: RETURN element formulation error
• IF Actor reference 無法 resolve 成 actor name: RETURN actor reference error

Phase 3 fail handling

• IF WRITE IO 失敗: RETURN IO error to caller
• IF target_path 已存在且 mode != overwrite: RETURN path conflict

Phase 4 fail handling

• IF validator exit != 0: RETURN syntax validation report to caller
• IF validator JSON parse 失敗: RETURN validator protocol error

Phase 5 fail handling

• IF RETURN 失敗（caller 已斷線）: WRITE ${$$target_path}.report ← $$report

§4 CROSS-REFERENCES

• 由上游 /aibdd-flows-specify（flow／activity modeling）DELEGATE。
• 上游必須先完成 Activity modeling；本 skill 不補需求、不改顆粒度、不新增 model element。