Skip to content

Conversation

Copy link
Contributor

Copilot AI commented Oct 22, 2025

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:
ActionsDictation Prompt GeneratorRun 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.

Copilot AI and others added 2 commits October 22, 2025 17:55
Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
Copilot AI changed the title [WIP] Add dictation prompt generation workflow Add dictation prompt generator agentic workflow Oct 22, 2025
Copilot AI requested a review from pelikhan October 22, 2025 18:04
Copy link
Contributor

@pelikhan pelikhan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@copilot schedule the workflow weekly

import reports.md

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
Copy link
Contributor Author

Copilot AI commented Oct 22, 2025

@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.

Copilot AI requested a review from pelikhan October 22, 2025 18:15
@pelikhan pelikhan marked this pull request as ready for review October 22, 2025 18:44
@pelikhan pelikhan merged commit 9e08545 into main Oct 22, 2025
5 checks passed
@pelikhan pelikhan deleted the copilot/create-dictation-prompt-workflow branch October 22, 2025 18:44
@github-actions
Copy link
Contributor

Agentic Changeset Generator triggered by this pull request.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants