Add dictation prompt generator agentic workflow

[Copilot]

Overview

This PR adds a new agentic workflow that generates comprehensive dictation instructions for optimizing speech-to-text interaction with the gh-aw repository. When triggered, the workflow analyzes the codebase to create a specialized prompt that helps convert casual dictation into precise, actionable agentic workflow instructions.

What's New

Workflow: dictation-prompt.md

A new workflow that automatically:

1. Scans the repository for project-specific terminology
2. Generates a glossary of approximately 100 key terms (95-105 acceptable)
3. Creates dictation instructions at .github/instructions/dictation.instructions.md
4. Provides guidelines for fixing common speech-to-text misrecognitions
5. Includes examples of transforming casual speech into professional agentic prompts
6. Creates a pull request with the generated file

Configuration

• Triggers:
  • workflow_dispatch (manual trigger)
  • schedule (weekly on Sundays at 6 AM UTC)
• Engine: copilot (GitHub Copilot CLI)
• Tools: edit, bash (full access for analysis), github (limited API access)
• Safe Outputs: create-pull-request (automatic PR creation)
• Permissions: Minimal (contents:read, actions:read)
• Timeout: 10 minutes
• Imports: shared/reporting.md (for standardized report formatting)

Use Case

The workflow addresses a common challenge when using speech-to-text to interact with the repository:

Problem: Speech recognition produces ambiguous or incorrectly transcribed technical terms:

"ghaw" instead of "gh-aw"
"work flow" instead of "workflow"
"lock file" instead of "lockfile" or ".lock.yml"
"MCP server thing" instead of "MCP server configuration"

Solution: The generated dictation instructions provide:

• A glossary of project-specific terms for recognition
• Guidelines for fixing ambiguous transcriptions
• Examples of converting casual speech to professional prompts

Example Transformations

The workflow teaches how to transform dictated input into proper agentic prompts:

Before (dictated):

"so I want you to maybe go ahead and compile the workflows and then after that you should probably check if there are any errors"

After (processed with dictation instructions):

"Compile all workflows in .github/workflows/ and validate for errors. Report any compilation failures with specific error messages."

Before (dictated):

"can you check why the MCP server thing isn't working in the copilot engine I think there might be something wrong with the configuration"

After (processed with dictation instructions):

"Investigate MCP server configuration issues in the copilot engine implementation. Check pkg/workflow/copilot_engine.go for MCP config generation. Review mcp-config.json format and verify environment variable setup."

How to Use

Trigger the workflow manually:

gh aw run dictation-prompt

Or via GitHub UI:
Actions → Dictation Prompt Generator → Run workflow

The workflow also runs automatically every Sunday at 6 AM UTC to keep the dictation instructions up to date with the latest repository changes.

The workflow will analyze the repository, generate the dictation instructions, and create a PR with the file at .github/instructions/dictation.instructions.md.

Testing

• ✅ Workflow compiles successfully (212.8 KB)
• ✅ All unit tests pass
• ✅ No compilation warnings or errors
• ✅ Structure matches existing workflow patterns
• ✅ Code review completed and feedback addressed
• ✅ Weekly schedule configured correctly

Files Added

• .github/workflows/dictation-prompt.md - Main workflow definition
• .github/workflows/dictation-prompt.lock.yml - Compiled GitHub Actions workflow

Original prompt

Create an agentic workflow that runs on workflow dispatch and generates a dictation prompt.

The prompt is designed to rephrase text generated by text-to-speech recognition and optimize it for the context of this repository.

• scan the repository and extract a 100 words glossary specific to the project, on top of known technical terms for software engineering. Provide glossary in prompt (no descriptions)
• fix ambiguous sentences by correcting the terminology to use the glossary or technical/software development.
• rephrase the tone to create a planning Agentic prompt that will be used on GitHub copilot agentic tasks.

Write results in a pull request to .github/instructions/dictation.instructions.md

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[pelikhan]

@copilot schedule the workflow weekly

import reports.md

[Copilot]

@copilot schedule the workflow weekly

import reports.md

Done in 70d89cc. The workflow now runs weekly on Sundays at 6 AM UTC and imports shared/reporting.md for report formatting guidelines.

[github-actions]

Agentic Changeset Generator triggered by this pull request.