dryforge
Agents fail in two ways — guessing what you meant, or optimizing for the checklist instead of the work.
A workflow harness plugin for Claude Code and Codex built around 'bounded autonomy': a ready step validates your intent and surfaces load-bearing unknowns before a spec is written, a go step executes the approved dependency graph instead of re-planning, and verification requires captured evidence (commands, diffs, runtime smoke) rather than the agent's self-report. Also maintains durable project memory across sessions.
- Validate intent and surface unknowns with ready before the agent writes a spec
- Execute an approved plan as a dependency graph with go
- Require captured evidence (commands, diffs, runtime smoke) before accepting work as done
Agents fail in two ways — guessing what you meant, or optimizing for the checklist instead of the work. dryforge fixes a minimum structure (validated intent, execution graph, evidence floor, project memory) while leaving the model's reasoning space open inside it.
Claude Code and Codex users who want a structured intent-plan-execute-verify workflow with durable project memory instead of ad-hoc prompting.
https://github.com/fn-opt/dryforge
By fn-opt
How to Get It
/plugin marketplace add fn-opt/dryforge (then: /plugin install dryforge)
Tip: Paste this into a Claude Code conversation. Verify command matches your Claude Code version.
After installing, paste this into Claude:
Help me validate intent and surface unknowns with ready before the agent writes a spec
Trust Signals Auto-scanned
Reviewer notes
Auto-scanned review. These are observations, not a security certification.
Scored from trust signals (evidence-eval-v1): 65 GitHub stars; contributors unknown; last commit 2d ago; license MIT.
Things to check
- Scanned, not hands-on tested — this entry was auto-scanned from public metadata (GitHub metrics, license, security flags). No reviewer has run it, and no tool-specific limitations have been documented yet.
How to evaluate tools before deploying →
Data shown here comes from public APIs and automated scanning. Reviewer notes reflect one person's experience. This is not a security certification or legal recommendation. Always evaluate tools according to your own organization's policies.
Evaluation
Scored from trust signals (evidence-eval-v1): 65 GitHub stars; contributors unknown; last commit 2d ago; license MIT.