From ai-writing-plugin
中文优先指导 generic_document mode 和 external document_profile.yaml use,同时保留 L1/L2/L3 boundaries、Markdown Spec candidate-profile rules、provenance、HITL、sample/reference 和 final-report limits。
How this skill is triggered — by the user, by Claude, or both
Slash command
/ai-writing-plugin:generic_documentThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill for `task_type: generic_document` and for explaining external profile runs such as `custom_technical_note`.
Use this skill for task_type: generic_document and for explaining external profile runs such as custom_technical_note.
默认用中文解释 generic_document workflow、external profile 使用方式、材料角色、critical claims、open confirmations 和 candidate updates。保留 generic_document、document_profile.yaml、Markdown Spec、candidate profile、profile-from-spec、custom_technical_note、HITL、NEEDS_USER_CONFIRMATION 等英文关键术语。
如果用户材料或 profile 字段为英文,可以保留字段名和引用片段;说明性文字优先中文。不要把 generic mode 或 external profile 描述成 official L3,也不要把 profile validation 写成专业批准。
generic_document supports evidence-aware professional document assistance when the user provides source, template, checklist, sample, reference, task, or profile information but there is no official L3 built-in document type.
It helps run the same plugin workflow and produces a review-ready package with open confirmations. It does not promise complete domain professional judgment.
generic_document is L1 generic mode, not an official L3 professional document type.
External document_profile.yaml is an L2 / customer profile mechanism. It must pass validation before use.
custom_technical_note is an external profile demo, not L3 and not an official built-in document type.
Typical generic_document inputs include:
sourcetemplatechecklistreferencesampledocument_profile.yamlThe generic_document default sections are:
External profiles may provide sections after validation, but they must still use the shared pipeline and artifact tree.
generic_document critical claims can come from task YAML, built-in generic rules, a validated external profile, or a spec-derived candidate profile after review.
Default generic critical claims include:
Critical claims are not automatically confirmed. They need source evidence or HITL, otherwise keep NEEDS_USER_CONFIRMATION, pending status, or an open item.
Requires human confirmation includes:
The engine must not convert external profile guidance into automatic professional approval.
Without sufficient T0/T1 support and explicit HITL, do not write final claims such as:
These are warning examples, not recommended output.
sample is not fact source. sample must not supply project facts, decisions, approvals, costs, schedules, compliance status, readiness, or risk acceptance. reference cannot prove project facts. reference is not project-specific fact support. T3/T4/T5 cannot support critical claim by themselves.
Artifacts should preserve provenance, source tier, claim status, evidence status, human confirmation status, profile id, and profile version when available.
Review focus includes:
Verification focus includes:
The generic_document final report is a review-ready package, not approval. Default final status should be ready_for_human_review, finalized_with_open_items, or blocked_pending_confirmation.
Open confirmations must remain visible when evidence or HITL is missing.
External profile runs use validated document_profile.yaml. Invalid profiles must fail safely and must not become a successful final report.
Markdown Spec is a human-readable upstream explanation layer. It can be used by profile-from-spec to generate a candidate profile, but Markdown Spec is not a runtime rule and cannot directly replace the structured profile or built-in rules.
Candidate profile output from profile-from-spec is proposed / inactive and cannot overwrite an active profile.
generic_document demo command:
.venv/bin/python -m ai_writing_plugin write-run --task examples/generic_document_demo_fixture/task.yaml
custom_technical_note external profile demo command:
.venv/bin/python -m ai_writing_plugin write-run --task examples/custom_technical_note_profile_demo_fixture/task.yaml
candidate_profile_update.yaml, candidate_skill_patch.md, and candidate profile outputs must not automatically replace stable Skill files or active profiles.Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub exquisite0828/ai_writer_plugin --plugin ai-writing-plugin