@fission-ai/openspec/dist/core/templates/workflows/feedback.js

AI-native system for spec-driven development

export function getFeedbackSkillTemplate() {
    return {
        name: 'feedback',
        description: 'Collect and submit user feedback about OpenSpec with context enrichment and anonymization.',
        instructions: `Help the user submit feedback about OpenSpec.

**Goal**: Guide the user through collecting, enriching, and submitting feedback while ensuring privacy through anonymization.

**Process**

1. **Gather context from the conversation**
   - Review recent conversation history for context
   - Identify what task was being performed
   - Note what worked well or poorly
   - Capture specific friction points or praise

2. **Draft enriched feedback**
   - Create a clear, descriptive title (single sentence, no "Feedback:" prefix needed)
   - Write a body that includes:
     - What the user was trying to do
     - What happened (good or bad)
     - Relevant context from the conversation
     - Any specific suggestions or requests

3. **Anonymize sensitive information**
   - Replace file paths with \`<path>\` or generic descriptions
   - Replace API keys, tokens, secrets with \`<redacted>\`
   - Replace company/organization names with \`<company>\`
   - Replace personal names with \`<user>\`
   - Replace specific URLs with \`<url>\` unless public/relevant
   - Keep technical details that help understand the issue

4. **Present draft for approval**
   - Show the complete draft to the user
   - Display both title and body clearly
   - Ask for explicit approval before submitting
   - Allow the user to request modifications

5. **Submit on confirmation**
   - Use the \`openspec feedback\` command to submit
   - Format: \`openspec feedback "title" --body "body content"\`
   - The command will automatically add metadata (version, platform, timestamp)

**Example Draft**

\`\`\`
Title: Error handling in artifact workflow needs improvement

Body:
I was working on creating a new change and encountered an issue with
the artifact workflow. When I tried to continue after creating the
proposal, the system didn't clearly indicate that I needed to complete
the specs first.

Suggestion: Add clearer error messages that explain dependency chains
in the artifact workflow. Something like "Cannot create design.md
because specs are not complete (0/2 done)."

Context: Using the spec-driven schema with <path>/my-project
\`\`\`

**Anonymization Examples**

Before:
\`\`\`
Working on /Users/john/mycompany/auth-service/src/oauth.ts
Failed with API key: sk_live_abc123xyz
Working at Acme Corp
\`\`\`

After:
\`\`\`
Working on <path>/oauth.ts
Failed with API key: <redacted>
Working at <company>
\`\`\`

**Guardrails**

- MUST show complete draft before submitting
- MUST ask for explicit approval
- MUST anonymize sensitive information
- ALLOW user to modify draft before submitting
- DO NOT submit without user confirmation
- DO include relevant technical context
- DO keep conversation-specific insights

**User Confirmation Required**

Always ask:
\`\`\`
Here's the feedback I've drafted:

Title: [title]

Body:
[body]

Does this look good? I can modify it if you'd like, or submit it as-is.
\`\`\`

Only proceed with submission after user confirms.`,
        license: 'MIT',
        compatibility: 'Requires openspec CLI.',
        metadata: { author: 'openspec', version: '1.0' },
    };
}
//# sourceMappingURL=feedback.js.map