Skip to content

feat: skip plan mode for fully-specified sequential tasks#44

Merged
barkain merged 2 commits into
mainfrom
fix/skip-plan-for-specified-tasks
Apr 12, 2026
Merged

feat: skip plan mode for fully-specified sequential tasks#44
barkain merged 2 commits into
mainfrom
fix/skip-plan-for-specified-tasks

Conversation

@barkain

@barkain barkain commented Apr 12, 2026

Copy link
Copy Markdown
Owner

Summary

  • Adds Step 3: Fully-Specified Task Detection to the routing table in commands/delegate.md
  • When the user provides explicit, ordered steps with no ambiguities, the orchestrator routes to direct sequential execution instead of entering plan mode unnecessarily
  • The route decision table becomes a 4-step check (was 3-step), with the new step inserted before the final route decision

Problem

The orchestrator entered plan mode for every multi-step task, even trivially specified ones like "bump version, commit, push, merge, tag, release." The user had to reject the plan and say "just do it." (#43)

Detection Indicators

  • Numbered/ordered steps in the request
  • Imperative sequential chains with concrete steps
  • Explicit signals: "just do it", "no plan needed", "skip planning"
  • All steps map trivially to tool calls
  • Zero ambiguity markers (no "should we", "which approach", "evaluate")

Test plan

  • Invoke /workflow-orchestrator:delegate bump version to X, commit, push, merge — should skip plan mode
  • Invoke /workflow-orchestrator:delegate create a calculator with tests and verify — should still enter plan mode (needs decomposition)
  • Invoke /workflow-orchestrator:delegate just do it: edit file X, commit, push — should detect "just do it" signal and skip plan mode

Closes #43

🤖 Generated with Claude Code

Add Step 3 (Fully-Specified Task Detection) to the routing table in
delegate.md. When the user provides explicit ordered steps with no
ambiguities or design decisions, the orchestrator now routes to direct
sequential execution instead of entering plan mode unnecessarily.

Detection indicators:
- Numbered/ordered steps in the request
- Imperative sequential chains with concrete steps
- Explicit signals: "just do it", "no plan needed"
- All steps map trivially to tool calls
- Zero ambiguity markers

Closes #43

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@qodo-code-review

Copy link
Copy Markdown
Contributor

Review Summary by Qodo

Add fully-specified task detection to skip plan mode

✨ Enhancement

Grey Divider

Walkthroughs

Description
• Adds Step 3 (Fully-Specified Task Detection) to routing table
• Detects explicit ordered steps to skip plan mode
• Enables direct sequential execution for trivial workflows
• Reduces unnecessary planning for concrete multi-step tasks
Diagram
flowchart LR
  A["User Request"] --> B["Step 1: Write Detection"]
  B --> C["Step 2: Breadth Detection"]
  C --> D["Step 3: Fully-Specified Task Detection"]
  D --> E{Task Type}
  E -->|Explicit Ordered Steps| F["Direct Sequential Execution"]
  E -->|Multi-phase Workflow| G["Plan Mode"]
  E -->|Read-only Breadth| H["Parallel Explore Agents"]
  E -->|Single Task| I["General-purpose Agent"]
Loading

Grey Divider

File Changes

1. commands/delegate.md ✨ Enhancement +17/-3

Add fully-specified task detection routing step

• Updated routing check from three-step to four-step process
• Added Step 3 (Fully-Specified Task Detection) with detection indicators
• Defined direct sequential execution route for explicit ordered steps
• Updated routing decision table with new fully-specified task pattern

commands/delegate.md


Grey Divider

Qodo Logo

@qodo-code-review

qodo-code-review Bot commented Apr 12, 2026

Copy link
Copy Markdown
Contributor

Code Review by Qodo

🐞 Bugs (3)   📘 Rule violations (0)   📎 Requirement gaps (0)   🎨 UX Issues (0)
🐞\ ≡ Correctness (1) ☼ Reliability (1) ⚙ Maintainability (1)

Grey Divider


Action required

1. Overbroad skip-plan criteria 🐞
Description
Step 3 claims fully-specified tasks have “no ambiguities”, but the routing trigger is “ANY of these”
indicators (including just “numbered steps”), so ambiguous numbered multi-step requests can be
incorrectly routed to DIRECT SEQUENTIAL EXECUTION and skip required planning/decomposition.
Code

commands/delegate.md[R41-50]

+**Pattern:** The user provided explicit, ordered steps with no ambiguities or design decisions needed. Every step is directly actionable without decomposition.
+
+**Indicators (ANY of these):**
+- Numbered or ordered steps in the request ("1. do X, 2. do Y, 3. do Z")
+- Imperative sequential chain ("do X, then Y, then Z" where each step is concrete)
+- Explicit skip-planning signals: "just do it", "no plan needed", "skip planning", "don't plan"
+- All steps map trivially to tool calls (edit file, run command, commit, push, merge, tag, release)
+- Zero ambiguity: no "should we", "which approach", "evaluate options", "design"
+
+**If fully-specified -> DIRECT SEQUENTIAL EXECUTION** (skip plan mode). Spawn one agent per step or batch related steps into a single agent. No `EnterPlanMode`, no `TaskCreate` during planning.
Evidence
The Step 3 pattern requires “no ambiguities”, but the checklist explicitly uses OR logic (“ANY of
these”), making a single weak signal (e.g., a numbered list) sufficient to skip plan mode even if
ambiguity markers are present elsewhere in the request.

commands/delegate.md[39-51]

Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The new “Fully-Specified Task Detection” uses an overly permissive OR condition ("ANY of these"), which can route ambiguous multi-step tasks to direct sequential execution.

## Issue Context
The text says the pattern requires “no ambiguities or design decisions”, but the detection criteria does not enforce that as a gate.

## Fix Focus Areas
- commands/delegate.md[39-51]
- commands/delegate.md[54-60]

## Suggested change
- Replace “Indicators (ANY of these)” with an explicit gated rule such as:
 - (skip-planning phrase present) OR (ordered steps AND each step is concrete)
 - AND no ambiguity markers are present ("should we", "which approach", "evaluate", "design", etc.)
 - AND steps map trivially to tool calls
- Add an explicit fallback: if any ambiguity marker exists, route to plan mode.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools



Remediation recommended

2. Prompt contradicts skip-plan routing 🐞
Description
The routing section instructs skipping plan mode for fully-specified sequential tasks, but later the
same system prompt states the main agent must not “skip plan mode” outside plan mode, creating
contradictory directives that can lead to inconsistent orchestration behavior.
Code

commands/delegate.md[50]

+**If fully-specified -> DIRECT SEQUENTIAL EXECUTION** (skip plan mode). Spawn one agent per step or batch related steps into a single agent. No `EnterPlanMode`, no `TaskCreate` during planning.
Evidence
Step 3 explicitly instructs “skip plan mode”, while the MAIN AGENT BEHAVIOR section later prohibits
skipping plan mode outside plan mode. With the new step, the prompt becomes self-inconsistent for
this newly introduced routing path.

commands/delegate.md[39-51]
commands/delegate.md[191-205]

Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The system prompt contains conflicting instructions: routing allows skipping plan mode for certain cases, but MAIN AGENT BEHAVIOR states the agent must not skip plan mode outside plan mode.

## Issue Context
This can cause the orchestrator to follow the wrong directive depending on which instruction the model prioritizes.

## Fix Focus Areas
- commands/delegate.md[39-51]
- commands/delegate.md[191-205]

## Suggested change
- Update MAIN AGENT BEHAVIOR to explicitly carve out the routing exceptions, e.g. “The main agent does NOT skip plan mode except when the mandatory routing check selects DIRECT EXECUTION / DIRECT SEQUENTIAL EXECUTION.”
- Ensure the terminology matches the routing table names (DIRECT EXECUTION vs DIRECT SEQUENTIAL EXECUTION).

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools



Advisory comments

3. Routing docs out of date 🐞
Description
The repo documentation and changelog still describe a 3-step routing flow (and “Step 3: Route
Decision”), but commands/delegate.md is now a 4-step flow; this inconsistency will mislead
maintainers/users about current behavior.
Code

commands/delegate.md[25]

+**Four-step routing check. MUST follow this order:**
Evidence
commands/delegate.md now states “Four-step routing check”, while multiple docs still describe
“3-step routing” and a “Step 3: Route Decision” flow, so the documentation set is inconsistent after
this PR.

commands/delegate.md[23-63]
docs/ARCHITECTURE_PHILOSOPHY.md[349-374]
docs/ARCHITECTURE_PHILOSOPHY.md[806-810]
docs/plan-explore-parallel-processing.md[1-5]
docs/plan-explore-parallel-processing.md[216-223]
CHANGELOG.md[206-213]

Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
Other documentation still references the old 3-step routing flow, but delegate.md is now 4-step.

## Issue Context
This is likely to confuse maintainers and anyone using ARCHITECTURE_PHILOSOPHY.md / plan-explore-parallel-processing.md / CHANGELOG as an authoritative description.

## Fix Focus Areas
- docs/ARCHITECTURE_PHILOSOPHY.md[349-374]
- docs/ARCHITECTURE_PHILOSOPHY.md[806-810]
- docs/plan-explore-parallel-processing.md[1-5]
- docs/plan-explore-parallel-processing.md[216-223]
- CHANGELOG.md[206-213]

## Suggested change
- Replace “3-step routing” references with “4-step routing” and update the step list/diagrams to include “Fully-Specified Task Detection” before the final route decision.
- Update any diagrams/ASCII boxes and arrow text accordingly.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools


Grey Divider

ⓘ The new review experience is currently in Beta. Learn more

Grey Divider

Qodo Logo

Comment thread commands/delegate.md
Comment on lines +41 to +50
**Pattern:** The user provided explicit, ordered steps with no ambiguities or design decisions needed. Every step is directly actionable without decomposition.

**Indicators (ANY of these):**
- Numbered or ordered steps in the request ("1. do X, 2. do Y, 3. do Z")
- Imperative sequential chain ("do X, then Y, then Z" where each step is concrete)
- Explicit skip-planning signals: "just do it", "no plan needed", "skip planning", "don't plan"
- All steps map trivially to tool calls (edit file, run command, commit, push, merge, tag, release)
- Zero ambiguity: no "should we", "which approach", "evaluate options", "design"

**If fully-specified -> DIRECT SEQUENTIAL EXECUTION** (skip plan mode). Spawn one agent per step or batch related steps into a single agent. No `EnterPlanMode`, no `TaskCreate` during planning.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Action required

1. Overbroad skip-plan criteria 🐞 Bug ≡ Correctness

Step 3 claims fully-specified tasks have “no ambiguities”, but the routing trigger is “ANY of these”
indicators (including just “numbered steps”), so ambiguous numbered multi-step requests can be
incorrectly routed to DIRECT SEQUENTIAL EXECUTION and skip required planning/decomposition.
Agent Prompt
## Issue description
The new “Fully-Specified Task Detection” uses an overly permissive OR condition ("ANY of these"), which can route ambiguous multi-step tasks to direct sequential execution.

## Issue Context
The text says the pattern requires “no ambiguities or design decisions”, but the detection criteria does not enforce that as a gate.

## Fix Focus Areas
- commands/delegate.md[39-51]
- commands/delegate.md[54-60]

## Suggested change
- Replace “Indicators (ANY of these)” with an explicit gated rule such as:
  - (skip-planning phrase present) OR (ordered steps AND each step is concrete)
  - AND no ambiguity markers are present ("should we", "which approach", "evaluate", "design", etc.)
  - AND steps map trivially to tool calls
- Add an explicit fallback: if any ambiguity marker exists, route to plan mode.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@barkain
barkain merged commit 8ac52d7 into main Apr 12, 2026
3 checks passed
@barkain
barkain deleted the fix/skip-plan-for-specified-tasks branch April 12, 2026 06:52
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.

Skip plan mode for fully-specified sequential tasks

1 participant