Add tools verification step to deepwork_jobs workflow #139
Conversation
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
nhorton
force-pushed
the
claude/update-deepwork-jobs-oGi2S
branch
2 times, most recently
from
14a8286 to
f9b074b
Compare
- Add new 'tools' step that verifies external tools are available - Tools step tests each required tool and finds alternatives if needed - Creates process-focused documentation (e.g., making_pdfs.md not pandoc.md) - Documents both installation and invocation for each process - Remove directory creation from implement step (define already does this) - Implement step now references tool documentation in generated instructions - Update quality criteria to include tool documentation references
- Update doc/architecture.md with new tools step in workflow - Fix outdated refine -> learn references in architecture doc - Update README.md refine -> learn command reference - Add changelog entry for tools step under [Unreleased]
- Add MCP servers and browser extensions as tool types - Remove documentation generation from examples - Add accessing_websites.md as example process doc - Separate installation docs into install_[tool].md files - Change quality criteria: "Alternatives Found" -> "Had Working Process" - Remove bash command examples (Claude can figure these out) - Generalize to cover CLI tools, MCPs, browser extensions, etc.
- Main agent analyzes job to determine required tools/capabilities - Spawns parallel sub-agents, one for each process requiring tooling - Each sub-agent tests, finds alternatives, and documents its process - Main agent reviews all sub-agent outputs for completeness - Updated quality criteria to reflect parallel sub-agent pattern
The sync command was displaying instructions to exit and restart Claude/Gemini sessions to use new skills. This is not actually needed, so remove the reload instructions output at the end of sync. Co-authored-by: Claude <noreply@anthropic.com>
…ques/ (#143) * feat(techniques): add .deepwork/techniques folder for reusable tool documentation - Add techniques folder created during install with AGENTS.md explaining format - Sync techniques to platform skill directories with dw_ prefix - Remove stale dw_ prefixed folders that no longer exist in techniques - Update tools step to create techniques instead of tools/ folder - Update learn step to create/refine techniques based on execution learnings - Update implement step to reference techniques instead of tools/ - Add comprehensive tests for technique sync functionality - Update architecture docs with techniques system documentation Techniques follow the Claude Skills format and are synced to all configured platforms, allowing reuse across multiple jobs. The dw_ prefix distinguishes DeepWork-managed techniques from user-created skills. BREAKING CHANGE: The tools step now creates techniques in .deepwork/techniques/ instead of .deepwork/jobs/[job_name]/tools/ * refactor(techniques): change prefix from dw_ to dwt_ Update the technique folder prefix in platform skill directories from "dw_" to "dwt_" for clearer identification of DeepWork techniques. * fix(techniques): update remaining dw_ references to dwt_ Fixed changelog entry and documentation that still referenced the old dw_ prefix instead of the new dwt_ prefix. * fix(techniques): update remaining dw_ references to dwt_ in code and tests - Update docstring in sync_techniques_to_platform to reference dwt_ prefix - Rename existing_dw_folders variable to existing_dwt_folders - Update comments in sync.py to reference dwt_ prefix - Update test docstrings and method names to use dwt_ terminology * fix(techniques): address code review findings Critical fixes: - Fix Gemini TOML format: use `prompt = """` instead of `[prompt]\ncontent = """` - Remove `name` field from Gemini TOML output (not part of Gemini format) - Add TOML string escaping for backslashes and quotes in description Minor fixes: - Fix TOCTOU race condition in stale folder removal using try/except - Update CHANGELOG to reference correct version v0.11.0 - Fix Gemini invocation syntax in AGENTS.md (use `/` not `:`) Tests: - Update existing tests to match new Gemini TOML format - Add test for TOML escaping with special characters --------- Co-authored-by: Claude <noreply@anthropic.com>
…ques/ - Add tools step to new_job workflow for verifying and documenting techniques - Techniques follow Claude Skills format with SKILL.md files - Spawn parallel sub-agents per technique for concurrent verification - Implement step now references techniques instead of inline tool docs - Add AGENTS.md documenting the techniques system - Regenerate skills to incorporate workflow-aware step progress
When no existing capability is available for a technique, the tools step now instructs sub-agents to prefer open source and free tools by default. Paid/proprietary tools should only be suggested if explicitly requested or no viable free alternative exists.
Includes fixes from main: - WebSearch permission - Skill template whitespace fix
nhorton
force-pushed
the
claude/update-deepwork-jobs-oGi2S
branch
from
97b965d to
95788f7
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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
Introduces a new
toolsstep (Step 3/5) to the deepwork_jobs workflow that verifies external tool availability and creates process-focused documentation before implementation. This ensures jobs can be successfully implemented by identifying tool requirements early and documenting how to use them.Key Changes
New
deepwork_jobs.toolsskill: Verifies required external tools are available, tests them, finds alternatives if needed, and creates process-focused documentation in.deepwork/jobs/[job_name]/tools/Workflow restructuring:
toolsstep betweenreview_job_spec(Step 2) andimplement(now Step 4)define→review_job_spec→tools→implement→learnImplement step refactoring:
definestep)Settings and permissions: Added necessary permissions to
.claude/settings.jsonfor the new skill to read/edit/write.deepwork/**and execute deepwork commandsDocumentation updates:
Implementation Details
The
toolsstep follows a structured approach:making_pdfs.mdnotpandoc.md)This separation of concerns ensures that tool verification happens before implementation begins, reducing implementation failures and creating reusable tool documentation for future jobs.