Skill: generate-questionnaire

Generates structured questionnaires for IT project management. Two types: (1) client questionnaire — collects initial requirements at project start; (2) team questionnaire — collects lessons learned at project closure. Also structures received answers for downstream skills.

Triggers

Russian: «сформируй опросник», «подготовь опросник для заказчика», «создай бриф заказчику», «подготовь анкету по урокам», «структурируй ответы заказчика», «нужен опросник», «подготовь вопросы для заказчика» English: "generate questionnaire", "create client brief", "prepare lessons questionnaire", "structure stakeholder answers", "build client brief", "prepare project brief"

Language Detection

Determine the language of the user's request:

If the request is in Russian → use templates with the -ru suffix
Otherwise → use templates with the -en suffix
All output (headings, labels, agent messages) must match the detected language.

Input

Data	Required	Source	Notes
Questionnaire type	yes	command or CLAUDE.md	`client` (phase 1) or `lessons` (phase 6)
Project name	yes	CLAUDE.md or command	Used in the file header
Recipient(s)	yes	command or CLAUDE.md	Name and email. For `client` — project sponsor; for `lessons` — team members
PM name	no	CLAUDE.md	Inserted into questionnaire metadata
Response deadline (`deadline`)	no	command	For `lessons` type only. If not provided — ask: "Specify the deadline by which the team should submit their responses."

If the type is not explicitly specified — determine from the project phase (from CLAUDE.md):

Phase 1 (initiation) → client
Phase 6 (closure) → lessons
Otherwise → ask one clarifying question:

Please clarify what you need:
A) Create a client questionnaire (Phase 1 — Initiation)
B) Create a team lessons-learned questionnaire (Phase 6 — Closure)
C) Structure received client answers
D) Structure received team answers

Output

Type	File	Project path	Mode
client	`questionnaire-client.md`	`{project}/01-initiation/`	generate
lessons	`questionnaire-lessons.md`	`{project}/04-closure/`	generate
any	Structured report in chat	—	structure

Instructions

Step 1 — Determine Mode and Type

Check the request type first:

If the request contains "evaluate", "analyze", "is it worth doing", "give a recommendation on the project" → this is an analysis request outside the skill's scope. Respond: "This skill structures questionnaire data but does not analyze answers or assess project viability. I can structure the answers and pass them to generate-charter, where the agent will flag ambiguities. Continue?" Wait for confirmation. Do not proceed until confirmed.

Determine the mode:

Command to generate / prepare / create a questionnaire → generate mode
Received answers ready to be structured → structure mode

If the mode is unclear — ask the clarifying question (see template above).

Step 2 — `generate` Mode: Create Questionnaire

Determine language.
Read the template:
- client → questionnaire-client-{lang}.md
- lessons → questionnaire-lessons-{lang}.md
Fill in metadata placeholders:

Placeholder	Source	Notes
`{{project_name}}`	CLAUDE.md or command
`{{recipient_name}}`	command or CLAUDE.md
`{{pm_name}}`	CLAUDE.md (if available)
`{{date}}`	current date (YYYY-MM-DD)
`{{deadline}}`	command	`lessons` type only. If not in command — ask before saving.

Save the file according to the Output table. 4a. If type is lessons and only 1 recipient is specified — append to the confirmation message: "1 recipient specified. If you need to add more — let me know before sending."
Show in chat:
- What was created: "Questionnaire saved as [filename] in [path]."
- Next step: "Prepare an email draft for sending?"
After confirmation — proceed to Step 3.
Log to logs/log.md: date, skill, created file, recipient.

Step 3 — Prepare Email and Call send-email

Determine parameters by type:

Type	Email template	Attachment	Format	Level
client	`email-tpl-questionnaire-client-{lang}.md`	questionnaire-client.md	.docx	3 (with confirmation)
lessons	`email-tpl-questionnaire-lessons-{lang}.md`	questionnaire-lessons.md	.md	2 (PM in CC)

Autonomy levels:

Level 2 — agent sends the email; PM added to CC automatically.
Level 3 — agent shows email draft; waits for human confirmation before sending.

Call send-email:

send-email(
  type          = "questionnaire-[client|lessons]",
  recipients    = [list of recipients with email],
  template      = [path to email template],
  attachment    = [path to questionnaire file],
  attach_format = [.docx | .md],
  level         = [3 | 2]
)

Stub: send-email is not yet implemented. Until it is — display an email draft based on the email template and offer to send manually via Gmail.

Step 4 — `structure` Mode: Structure Answers

Receive answers from chat or an attached file.
Determine the answer type:
- Answers about project goals, scope, constraints → client type
- Answers about lessons learned and retrospective → lessons type
Structure by blocks matching the corresponding questionnaire template.
Flag unanswered and ambiguous questions.
Show in chat:
- Structured data by block
- List of unanswered / ambiguous items (if any)
- Recommendation: "Pass data to [generate-charter | generate-closure-report]?"
On confirmation — offer to launch the downstream skill.

Dependencies

Reads:

CLAUDE.md of the managed project — project name, PM, contacts
Templates from templates/: questionnaires (4 files) + email templates (4 files)

Passes data to:

generate-charter — after structuring client answers (phase 1)
generate-closure-report — after structuring lessons answers (phase 6)

Calls:

send-email — delivery module (stub until implemented)

Constraints

Questionnaire template is fixed — questions are not generated dynamically by project type
Does not analyze answers or assess project viability
Does not send email directly — calls send-email
Does not create the project charter or close the project — only passes data downstream
Handles one project per invocation

Changelog

Date	Version	Change
2026-04-06	1.0	Skill created. Two questionnaire types (client, lessons). Two modes (generate, structure). send-email stub.
2026-04-06	1.1	Post-eval fixes (4/6 → 6/6): added `{{deadline}}` to input and placeholder tables; added step 4a for single-recipient warning (lessons); added guard rail in Step 1 for analysis requests.

generate-questionnaire