Skip to content

fix: make agent prompts imperative about initiative delegation workflow#20

Merged
danielkov merged 3 commits intomainfrom
danielkov/fix/improve-agent-subcommand-augmentation
Feb 22, 2026
Merged

fix: make agent prompts imperative about initiative delegation workflow#20
danielkov merged 3 commits intomainfrom
danielkov/fix/improve-agent-subcommand-augmentation

Conversation

@danielkov
Copy link
Copy Markdown
Collaborator

@danielkov danielkov commented Feb 22, 2026

Summary

  • Agents planning multi-project initiatives were creating tasks directly instead of delegating to sub-agents
  • Redesigned all agent-facing prompts to front-load constraints and distinguish single-project vs multi-project paths
  • granary initiate now opens with a CRITICAL: Delegation-Only Workflow section that explicitly forbids direct task creation
  • granary plan (without --project) now includes a scope guard warning initiative agents to stop and use initiate instead
  • Entrypoint and --help now present a decision tree ("Choose ONE entry point based on scope") instead of a flat command list

New granary initiate prompt (top section)

## CRITICAL: Delegation-Only Workflow

You are the initiative coordinator. Your job is to create projects and delegate.
Do NOT create tasks directly. Do NOT use `granary project <id> tasks create`.
Task creation is handled by sub-agents via `granary plan --project <id>`.

## Workflow

1. Create projects with descriptions (Step 1 below)
2. Set inter-project dependencies (Step 2 below)
3. Delegate planning to sub-agents: `granary plan --project <id>` for each project
4. Verify the initiative

New granary plan scope guard

## Scope: Single-Project Planning

This workflow is for planning ONE project with tasks.
If this is part of a multi-project initiative, stop here.
Use `granary initiate "Initiative name"` instead — it guides you through
creating projects and delegating task planning to sub-agents.

New entrypoint prompt

Available granary workflows:

Choose ONE entry point based on scope:
- Single project (one feature/fix): `granary plan "Feature name"`
- Multiple projects (cross-cutting work): `granary initiate "Initiative name"`

Other commands:
- Work on task: `granary work <task-id>`
- Search: `granary search "keyword"`

New --help text

AGENTS (AI/LLM):
  Choose ONE entry point based on scope:

  Single project (one feature/fix):
    granary plan "Feature name"

  Multiple projects (cross-cutting work):
    granary initiate "Initiative name"
    NOTE: Do NOT use `granary plan` for multi-project work.
    `initiate` guides you through creating projects and delegating task planning to sub-agents.

✻ Clauded...

danielkov and others added 3 commits February 22, 2026 17:43
Agents planning multi-project initiatives were creating tasks directly
instead of delegating to sub-agents. Redesigned prompts across initiate,
plan, entrypoint, and --help to front-load constraints and distinguish
single-project vs multi-project paths.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@danielkov danielkov merged commit 8253184 into main Feb 22, 2026
3 checks passed
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.

1 participant