
implementation
by Yuvasee
SKILL.md
name: implementation description: Execute implementation tasks with different approaches (single, dual-agent, plan-based).
Implementation
Executes implementation tasks using different approaches: direct execution, dual-agent comparison, or plan-based execution.
Requirements
- Active session must exist (session path in working memory)
- If no active session: STOP and ask user for session path
Default Action: dop2
IMPORTANT: For implementation phases, use dop2 by default.
-
dop2 (Dual-Agent with Auto-Selection): Default for most work
- Use when task has design decisions or tradeoffs
- Use when "clean" vs "minimal" approach might differ
- When in doubt, use dop2
-
dop (Direct Execution): Only for trivially simple changes
- Use for 1-2 line fixes with no design decisions
- Use for obvious mechanical changes (rename, move, etc.)
- Rare cases only
Rule of thumb: If you're considering dop, ask "Could there be a cleaner way?" If yes -> use dop2.
Actions
do - Direct Implementation
Execute a task directly with full implementation and documentation.
Session path: [SESSION_PATH from working memory] Task: $ARGUMENTS
Steps
-
Investigate and implement:
- Review session documents for context
- Explore codebase as needed
- Make all necessary code changes
- Run pyright/ruff (Python) or tsc (TypeScript) - fix all errors
-
Document work:
- Create
[SESSION_PATH]/[TIMESTAMP_FILE]-do-[task-slug].md:
# Task: [brief title] Date: [TIMESTAMP_LOG] ## What was done [Description of changes] ## Files Changed - [file] - [what changed] ## Testing [How verified] ## Notes [Any important observations] - Create
-
Update session:
- Edit
_overview.md:- Flow Log:
- [TIMESTAMP_ITERATION] Task: [title] -> [filename].md - Files:
- [filename].md - [brief description]
- Flow Log:
- Commit session (if git repo):
cd [SESSION_DIR] && git add . && git commit -m "Task: [title]"
- Edit
-
Git workflow:
- Read Working Dir from session
_overview.md - If Working Dir not set: check project
.samocodefile forMAIN_REPO, or use git root, or ask user - Check branch:
cd [WORKING_DIR] && git branch --show-current - If on feature branch: commit with short message
- If on main: warn user to create branch first
- Read Working Dir from session
-
Report back: Summary of what was done
do2 - Dual-Agent Comparison
Spawn two agents with different philosophies to solve the same task independently. Compare solutions and present options.
Session path: [SESSION_PATH from working memory] Task: $ARGUMENTS
Dual Agent Execution
Spawn 2 sub-agents in parallel. Do NOT tell them about each other.
Both agents solve the entire task independently with different philosophies.
Agent 1: Minimal Footprint
Session path: [SESSION_PATH] Task: $ARGUMENTS
Approach: Smallest possible change surface.
- Touch few files, prefer localized fixes
- Avoid new abstractions
- Work within existing structures
Deliverables:
-
Create
[SESSION_PATH]/[TIMESTAMP_FILE]-solution-minimal.md(use TIMESTAMP_FILE from Session Context):# Solution: Minimal Footprint Task: [description] ## Approach [High-level strategy] ## Proposed Changes ### File: [path] ```[language] [code]Why: [rationale]
Tradeoffs
[What we accept]
-
DO NOT edit actual code files - proposals only
-
Report summary
Agent 2: Clean Foundation
Session path: [SESSION_PATH] Task: $ARGUMENTS
Approach: The "right way" with time to do it properly.
- Refactor for cleaner solution
- Introduce helpful abstractions
- Consider future maintainability
Deliverables:
-
Create
[SESSION_PATH]/[TIMESTAMP_FILE]-solution-clean.md(use TIMESTAMP_FILE from Session Context):# Solution: Clean Foundation Task: [description] ## Approach [High-level strategy] ## Proposed Changes ### File: [path] ```[language] [code]Why: [rationale]
Architecture Improvements
[New patterns, benefits]
-
DO NOT edit actual code files - proposals only
-
Report summary
Synthesis (Parent does this)
After both agents complete:
-
Review both solutions
-
Create
[SESSION_PATH]/[TIMESTAMP_FILE]-comparison.md:# Comparison: [task] ## Minimal Footprint **Pros:** [list] **Cons:** [list] ## Clean Foundation **Pros:** [list] **Cons:** [list] ## Recommendation [Which approach and why] ## Hybrid Possibility [Can we combine best of both?] -
Update
_overview.md:- Flow Log:
- [TIMESTAMP_ITERATION] Dual analysis: [task] -> comparison.md - Files: all three solution docs
- Flow Log:
-
Commit session (if git repo):
cd [SESSION_DIR] && git add . && git commit -m "Dual analysis: [task]" -
Present to user:
Which approach? 1. Minimal footprint 2. Clean foundation 3. Hybrid 4. Neither (rethink)
After User Confirms
- Implement chosen solution
- Run pyright/ruff (Python) or tsc (TypeScript) - fix all errors
- If part of a plan: Update plan progress (mark items
- [x], add completion note) - Git workflow: Read Working Dir from session
_overview.md, ensure on feature branch, commit changes - Update session: Edit
_overview.mdwith Flow Log entry, commit
dop - Plan Phase Execution
Execute a specific phase from the plan without comparing approaches.
IMPORTANT: dop is for TRIVIALLY SIMPLE phases only (1-2 line changes). Use dop2 for everything else.
Session path: [SESSION_PATH from working memory] Phase: $ARGUMENTS
IMPORTANT: Only work on this specific part. Don't do other parts of the plan.
Steps
-
Find the plan:
- Check session
_overview.md-> Plans section for plan file - Read the plan file from session folder
- Locate the specific phase/section to implement
- Check session
-
Implement:
- Make all necessary code changes for this phase only
- Follow codebase best practices
- Run pyright/ruff (Python) or tsc (TypeScript) - fix all errors
-
Document work:
- Create
[SESSION_PATH]/[TIMESTAMP_FILE]-dop-[phase-slug].md:
# Phase: [phase name] Date: [TIMESTAMP_LOG] Plan: [plan filename] ## Completed Items - [x] [Item from plan] - [x] [Item from plan] ## Changes Made - [file] - [what changed] ## Testing [How verified] ## Notes [Any issues or observations] - Create
-
Update plan progress (MANDATORY):
- Read plan file from session folder
- For each completed item in this phase:
- Use Edit tool to change
- [ ]to- [x]
- Use Edit tool to change
- Add phase completion note:
**Phase completed** ([TIMESTAMP_ITERATION]) - Verify edits by reading plan file again
- Remove obsolete info, keep plan clean
-
Update session:
- Edit
_overview.md:- Flow Log:
- [TIMESTAMP_ITERATION] Phase done: [phase] -> [filename].md - Files:
- [filename].md - [phase description]
- Flow Log:
- Commit session (if git repo):
cd [SESSION_DIR] && git add . && git commit -m "Phase: [phase]"
- Edit
-
Git workflow:
- Read Working Dir from session
_overview.md - If not set: check project
.samocodefile forMAIN_REPO, or use git root, or ask user - Check branch:
cd [WORKING_DIR] && git branch --show-current - If on feature branch: commit with short message
- If on main: warn user
- Read Working Dir from session
-
Report back: Summary of completed items
dop2 - Dual-Agent Plan Phase with Auto-Selection
DEFAULT IMPLEMENTATION STRATEGY - Use this for all non-trivial phases.
Execute a plan phase using dual-agent comparison, with automatic solution selection weighted toward clean approach.
Session path: [SESSION_PATH from working memory] Task: $ARGUMENTS
IMPORTANT: Only work on this specific task. Don't do other parts of the plan.
Phase Scope Assessment
Before spawning agents, assess phase complexity:
- Review the phase items - Count distinct deliverables/changes
- If phase has 5+ items or touches 5+ files:
- Consider splitting into sub-phases (e.g., "Phase 3a", "Phase 3b")
- Each sub-phase should be a coherent unit (1-3 related items)
- Run dop2 for each sub-phase sequentially
- If phase has 2-4 independent sub-items:
- Can run 2 dop2 calls in parallel for independent items
- Only parallelize if items don't share files/dependencies
Parallelization Rule: Run up to 2 dop2 calls in parallel for independent items:
- Each dop2 spawns its own pair (minimal + clean) = 4 sub-agents total
- Only parallelize if items don't share files/dependencies
- Items must be self-contained with no data dependencies
Dual Agent Execution
Spawn 2 sub-agents in parallel with the context below. Do NOT tell them about each other.
CRITICAL: Both agents solve the entire task independently. This is NOT task splitting - each agent produces a FULL solution with their own philosophy. We compare two complete approaches, not partial results to assemble.
Agent 1: Minimal Footprint
Session path: [SESSION_PATH] Task: $ARGUMENTS
Your approach: Solve with the smallest possible change surface.
Guidelines:
- Touch as few files as possible
- Prefer simple, localized fixes
- Avoid new abstractions or patterns
- Prioritize "getting it done" over elegance
- Work within existing structures
Deliverables:
-
Create
[SESSION_PATH]/[TIMESTAMP_FILE]-solution-minimal.md:# Solution: Minimal Footprint Date: [TIMESTAMP_LOG] Task: [description] ## Approach [High-level strategy] ## Proposed Changes ### File: [path] ```[language] [code]Why: [rationale]
Tradeoffs
[What we accept with this approach]
Risks
[Potential issues]
-
Update
_overview.md:- Flow Log:
- [TIMESTAMP_ITERATION] Solution (minimal): [task] -> [filename].md - Files:
- [filename].md - Minimal solution
- Flow Log:
-
DO NOT edit actual code files - suggestions only
-
Report back with summary
Agent 2: Clean Foundation
Session path: [SESSION_PATH] Task: $ARGUMENTS
Your approach: Solve the "right way" assuming time to do it properly.
Guidelines:
- Refactor if it leads to cleaner solution
- Introduce abstractions where they reduce future complexity
- Consider how this code will evolve
- Prioritize maintainability over minimal diff
- Fix adjacent issues if tightly related
Deliverables:
-
Create
[SESSION_PATH]/[TIMESTAMP_FILE]-solution-clean.md:# Solution: Clean Foundation Date: [TIMESTAMP_LOG] Task: [description] ## Approach [High-level strategy] ## Proposed Changes ### File: [path] ```[language] [code]Why: [rationale]
Architecture Improvements
[New patterns, refactorings]
Benefits
[Long-term advantages]
-
Update
_overview.md:- Flow Log:
- [TIMESTAMP_ITERATION] Solution (clean): [task] -> [filename].md - Files:
- [filename].md - Clean solution
- Flow Log:
-
DO NOT edit actual code files - suggestions only
-
Report back with summary
Synthesis & Auto-Selection
After both agents complete:
-
Create
[SESSION_PATH]/[TIMESTAMP_FILE]-comparison-[task-slug].md:# Comparison: [task] ## Key Differences [What's fundamentally different] ## Solution 1: Minimal **Pros:** [list] **Cons:** [list] **Risk:** [Low/Medium/High] ## Solution 2: Clean **Pros:** [list] **Cons:** [list] **Risk:** [Low/Medium/High] ## Recommendation [Which approach and why] ## Hybrid Possibility [Can we combine best of both?] -
Auto-Selection Logic:
- Default: Choose Clean Foundation unless minimal has compelling justification
- Choose Minimal only if:
- Clean introduces significant complexity with unclear benefit
- Minimal solves urgent issue with acceptable tradeoffs
- Clean approach has high risk/uncertainty
- Weight toward Clean means: when in doubt, prefer clean approach
-
Update
_overview.md:- Flow Log:
- [TIMESTAMP_ITERATION] Comparison: [task] -> [filename].md - Files:
- [filename].md - Solution comparison
- Flow Log:
-
Commit session (if git repo):
cd [SESSION_DIR] && git add . && git commit -m "Dual analysis: [task]" -
Present comparison and ask:
Which approach? 1. Minimal footprint 2. Clean foundation 3. Hybrid 4. Neither (rethink)
After User Confirms
- Implement chosen solution
- Run pyright/ruff (Python) or tsc (TypeScript) - fix all errors
- Update plan progress (MANDATORY):
- Read plan file from session folder (find via
_overview.mdPlans section) - For each completed item in this phase:
- Use Edit tool to change
- [ ]to- [x]
- Use Edit tool to change
- Add phase completion note:
**Phase completed** ([TIMESTAMP_ITERATION]) - Verify edits by reading plan file again
- Read plan file from session folder (find via
- Git workflow:
- Read Working Dir from session
_overview.md - Ensure on feature branch (not main)
- Commit code changes
- Read Working Dir from session
- Update session:
- Edit
_overview.mdwith Flow Log entry:- [TIMESTAMP_ITERATION] Phase implemented: [name] - Commit session changes
- Edit
IMPORTANT! If unsure about something, ask first.
Score
Total Score
Based on repository quality metrics
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
3ヶ月以内に更新がある
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
Reviews
Reviews coming soon