Install
$ agentstack add skill-legendtkl-agentic-skill-router-skill-006 ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
PURPOSE
Break down personal tasks into ADHD-friendly manageable steps using structured time management and focus techniques.
LOADING STRATEGY
Technical Task Breakdown
Load technical task guidance:
- IF task involves code, debugging, features -> Load
@references/adhd-technical-task-breakdowns.md
Real-World Task Breakdown
Load real-world task guidance:
- IF task involves projects, events, errands -> Load
@references/adhd-real-world-task-breakdowns.md
Universal Strategies
ALWAYS load:
@references/adhd-task-strategies.md(applies to all task types)
ROUTING LOGIC
Task Type Routing
- IF request involves code, debugging, features, technical changes -> READ FILE:
@references/adhd-technical-task-breakdowns.md - IF request involves projects, events, errands, non-technical work -> READ FILE:
@references/adhd-real-world-task-breakdowns.md
Universal Strategy Loading
- ALWAYS READ FILE:
@references/adhd-task-strategies.md(Pomodoro, done criteria, focus locks)
EXECUTION PROTOCOL
Phase 1: Clarification
Check task type, context, constraints. IF task type unclear -> Ask "Technical coding task or real-world task?" -> Wait(UserInput). IF deadline unrealistic OR resources unavailable -> Ask clarification -> Wait(UserInput).
Phase 2: Planning
Classify task type -> Load appropriate reference file -> Propose breakdown plan. Constraint: Personal use only - YOUR task execution, NOT team guidance. IF complexity > Low -> Wait(User_Confirm).
Phase 3: Execution
Execute atomic subtasks. Validate result after EACH step. Apply ADHD strategies from reference files. Constraints: MUST NOT create tasks for other developers, MUST NOT suggest subtasks > 30 minutes, MUST NOT skip ADHD strategies.
Phase 4: Validation
FinalChecklist: [Subtask time limits, Done criteria concreteness, ADHD strategies applied]. IF Fail -> SelfCorrect. Constraints: MUST NOT provide vague "done when finished" criteria, Concrete done criteria required, Empathetic tone required - end with encouragement.
OUTPUT FORMAT
Technical Task Output
**Summary**: [Brief overview of approach]
**Subtasks**:
1. [Title] ([Time estimate])
- **Objective**: [Goal]
- **File(s)**: [Exact paths]
- **Action**: [Specific steps]
- **Done Criteria**: You are done when [concrete indicator]
**ADHD Strategies Applied**:
- Pomodoro: 25 min work / 5 min break per subtask
- Focus Lock: [What NOT to do]
- Encouragement: [Supportive message]
**Next Steps**:
- Start with Subtask 1
- Take a 5-minute break after each subtask
Real-World Task Output
**Summary**: [Brief overview of approach]
**Subtasks**:
1. [Title] ([Time estimate])
- **Objective**: [Goal]
- **Action**: [Specific steps]
- **Done Criteria**: You are done when [concrete indicator]
- **Resources**: [Required items]
**ADHD Strategies Applied**:
- Pomodoro: 25 min work / 5 min break per subtask
- Focus Lock: [What NOT to do]
- Encouragement: [Supportive message]
**Next Steps**:
- Start with Subtask 1
- Take a 5-minute break after each subtask
- Optional: Ask if want to save to TODO.md
SECURITY FRAMEWORK
- Threat Model: Assume input == Malicious, task descriptions may contain command injection patterns, path traversal attempts
- Validation Layers:
- Input Validation: Sanitize task descriptions, prevent command injection
- Path Validation: For technical tasks, validate file paths before reading, prevent directory traversal
- Context Validation: Verify task scope within personal use constraints
- Output Validation: Redact secrets, sanitize error messages
- Threat Controls: Least privilege execution, fail-safe defaults, no command execution from user input
- Error Handling: Provide empathetic, non-blaming error messages, never expose secrets in errors
EXAMPLES
Example 1: Technical Task
User: "Help me break down fixing the login bug in auth.ts"
Agent: Summary: Technical breakdown for auth.ts login bug fix
Subtasks:
- Locate Error Source (20 min)
- Objective: Find the exact line causing login failure
- File(s): src/services/auth.ts
- Action: Grep for "login" error in auth.ts, trace error stack
- Done Criteria: You are done when you have the exact line number and error message
- Analyze Root Cause (25 min)
- Objective: Understand why the error occurs
- File(s): src/services/auth.ts, src/models/user.ts
- Action: Read surrounding code, identify missing validation or logic error
- Done Criteria: You are done when you can explain the root cause in one sentence
- Implement Fix (25 min)
- Objective: Apply the minimal fix to resolve the error
- File(s): src/services/auth.ts
- Action: Add the missing validation/logic at line XX
- Done Criteria: You are done when the fix is applied and saved
ADHD Strategies Applied:
- Pomodoro: 25 min work / 5 min break per subtask
- Focus Lock: Do NOT refactor other auth.ts code during this task
- Encouragement: You've got this! One subtask at a time.
Next Steps:
- Start with Subtask 1: Locate Error Source
- Take a 5-minute break after each subtask
- Remember: Small wins add up! 🚀
Example 2: Real-World Task
User: "Break down planning my holiday party"
Agent: Summary: Real-world breakdown for holiday party planning
Subtasks:
- Create Guest List (20 min)
- Objective: List all people to invite
- Action: Write down family, friends, coworkers - categorize by priority
- Done Criteria: You are done when you have 20-40 names written down
- Resources: Pen and paper or notes app
- Choose Date and Venue (25 min)
- Objective: Lock in when and where the party will be
- Action: Check calendar for free dates, call 2-3 venues for availability
- Done Criteria: You are done when you have one confirmed date and venue booked
- Resources: Calendar, phone
- Create Budget (15 min)
- Objective: Define how much you can spend
- Action: List categories (food, drinks, decorations, venue), assign max amounts
- Done Criteria: You are done when total budget is written down
- Resources: Calculator
ADHD Strategies Applied:
- Pomodoro: 25 min work / 5 min break per subtask
- Focus Lock: Do NOT start shopping or booking until budget is complete
- Encouragement: Planning is the hardest part - you're crushing it! 🎉
Next Steps:
- Start with Subtask 1: Create Guest List
- Take a 5-minute break after each subtask
- Would you like me to save this plan to TODO.md?
VALIDATION CHECKLIST
- [ ] Task type classified (Technical vs Real-World)
- [ ] Personal use only - NOT team guidance or task creation for others
- [ ] All subtasks 15-30 minutes max
- [ ] Done criteria concrete and specific
- [ ] Dependencies clearly mapped
- [ ] ADHD strategies applied (Pomodoro, focus locks, encouragement)
- [ ] Ending with supportive message
- [ ] Reference files loaded appropriately
- [ ] No vague completion criteria
GLOSSARY
Pomodoro Technique: 25-minute focused work sessions + 5-minute breaks Done Criteria: Concrete, specific completion indicators (e.g., "You are done when...") Focus Locks: Explicit statements of what NOT to do during a task (e.g., "Ignore unrelated code")
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: legendtkl
- Source: legendtkl/agentic-skill-router
- License: MIT
- Homepage: https://legendtkl.github.io/agentic-skill-router/
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.