skills.rest
JaimeHoracio
JaimeHoracioCOMMUNITY
Explore all skills by this author →
GitHub

execution-mode-evaluation

Analyzes CodeGraph dependency data to recommend inline or subagent-driven task execution.

1|Updated Jun 6, 2026
Search Tags:#dependency analysis#impact analysis#subagent#task orchestration#openspec#codegraph#execution mode
One-click install
npx skills add https://github.com/JaimeHoracio/Ostacky --skill execution-mode-evaluation-jaimehoracio
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: execution-mode-evaluation
Source: https://github.com/JaimeHoracio/Ostacky/tree/main/assets/skills/execution-mode-evaluation
Command: npx skills add https://github.com/JaimeHoracio/Ostacky --skill execution-mode-evaluation-jaimehoracio

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When implementing a change with multiple tasks, it is unclear whether to execute everything inline in one context or dispatch work to parallel subagents. This Skill removes the guesswork by analyzing real dependency data from CodeGraph and applying strict precedence rules to recommend the optimal execution mode. ## Core Features & Use Cases - Dependency Mapping: Builds shared-file maps, file clusters, and sequential dependency lists from tasks.md and CodeGraph impact analysis. - Rule-Based Decision: Applies ordered global rules (single cluster, cross-cluster dependencies, task count, line estimates) plus per-phase evaluation to choose INLINE or SUBAGENT_DRIVEN. - Controller Contract: Produces a structured JSON snapshot with expectedTaskIds, clusters, and the triggered rule, gated by explicit user confirmation before persisting. - Use Case: Given 5 implementation tasks spanning 3 independent file clusters, the Skill recommends subagent-driven execution with one subagent per cluster running in parallel. ## Quick Start Analyze the tasks in the active OpenSpec change with CodeGraph and tell me whether to execute inline or with subagents.

Frequently Asked Questions about execution-mode-evaluation

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I decide between inline and subagent-driven execution for coding tasks?▼

Map which files each task modifies using CodeGraph, build file clusters from shared files, then apply ordered rules: one connected cluster means inline, independent clusters without sequential dependencies mean subagent-driven with one subagent per cluster.

What is the execution-mode-evaluation skill used for?▼

It analyzes implementation tasks from an OpenSpec change using CodeGraph dependency data and recommends INLINE or SUBAGENT_DRIVEN execution. The output is a JSON snapshot consumed by a controller after explicit user confirmation.

Does execution-mode-evaluation require CodeGraph to work?▼

CodeGraph is the primary data source, but the skill has fallbacks. If CodeGraph is unavailable it returns a low-confidence INLINE recommendation, and if a valid discovery-cache snapshot exists it reuses that data instead of calling CodeGraph again.

When should tasks run inline instead of in subagents?▼

Run inline when tasks share files in a single cluster, when sequential dependencies exist without an explicit contract, when there are fewer than 3 independent tasks, or when the estimated change is under 30 lines.

Why does the snapshot require expectedTaskIds?▼

The controller mandates expectedTaskIds whenever taskCount is greater than zero and rejects snapshots missing it. The only exception is an early-exit result for trivial changes with two or fewer tasks and no CodeGraph calls.

Related Skills

Explore More Software Engineering Skills→
joshuakesselljoshuakessell

code-graph

Index and query a persistent code graph for symbol lookup and blast-radius analysis.

Community
Intermediate
mcampbellrmcampbellr

gsd-fast

Execute simple one-sentence tasks inline without planning or subagent spawning.

Community
Basic
silverbee-aisilverbee-ai

session-cache

Cache previously fetched domain data within a conversation to avoid duplicate API calls.

Official
Intermediate
Yjj0333Yjj0333

codegraph-skill-router

Routes code exploration tasks between CodeGraph index queries and ordinary repository search.

Community
Intermediate