Update quality validation prompts to use natural language guidance#141
Merged
Update quality validation prompts to use natural language guidance#141
Conversation
Remove explicit JSON response format instructions from prompt-based hooks.
Instead of instructing to respond with {"ok": true} or {"ok": false, "reason": "..."},
use simpler language like "let the agent finish" or "have the agent keep working".
Updated:
- skill-job-step.md.jinja template
- hooks_system.md documentation
- Regenerated all skill files via deepwork install
* docs: add AGENTS.md for claude templates with prompt hook guidance Add notes for AI agents working on jinja templates, specifically documenting that prompt-based hooks should not include response format instructions since Claude Code's internal instructions already specify the expected format. Reference: anthropics/claude-code#11786 * Update AGENTS.md with clarification on response format Clarified instructions regarding response format in prompt-based hooks. --------- Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This PR updates the quality validation hook prompts across the codebase to use natural language guidance instead of JSON response formats. The changes make the validation system more intuitive by instructing agents to "let the agent finish" or "have the agent keep working" rather than requiring specific JSON responses.
Key Changes
Settings Configuration: Added deepwork file access and bash command permissions to
.claude/settings.jsonRead(./.deepwork/**),Edit(./.deepwork/**),Write(./.deepwork/**),Bash(deepwork:*)Quality Validation Prompts: Updated validation hook prompts across all skill definitions to use natural language instructions:
{"ok": true}→ to: "let the agent finish"{"ok": false, "reason": "..."}→ to: "have the agent keep working until all criteria are satisfied"deepwork_jobs.define/SKILL.mddeepwork_jobs.implement/SKILL.mddeepwork_jobs.learn/SKILL.mddeepwork_jobs.review_job_spec/SKILL.mdmanual_tests.infinite_block_tests/SKILL.mdmanual_tests.reset/SKILL.mdmanual_tests.run_fire_tests/SKILL.mdmanual_tests.run_not_fire_tests/SKILL.mdDocumentation: Updated hook system documentation in
doc/platforms/claude/hooks_system.mdto reflect the new natural language validation patternTemplates: Updated the skill job step template in
src/deepwork/templates/claude/skill-job-step.md.jinjato generate validation prompts with the new formatImplementation Details
The changes maintain the same validation logic but express it in a more conversational way that aligns better with how Claude naturally understands task completion. Instead of expecting structured JSON responses, the system now guides agents through natural language cues about whether to continue working or finish.