소스 정보
- 저장소
- diegosouzapw/awesome-omni-skill
- 최근 소스 활동
- 2026년 2월 28일 04:04
- 감지된 SKILL.md 언어
- 영어
- 스타
- 50
- 포크
- 19
설치 방법
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
소스 파일 검토
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
메뉴
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/diegosouzapw/awesome-omni-skill --skill session-start명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
Token-efficient tracking for AI orchestration. CLI-first for status updates (~50 tokens), agent fallback for complex ops (~1KB). Use when: updating task status, querying blockers, creating progress files, validating phases.
AshAi extension guidelines for integrating AI capabilities with Ash Framework. Use when implementing vectorization/embeddings, exposing Ash actions as LLM tools, creating prompt-backed actions, or setting up MCP servers. Covers semantic search, LangChain integration, and structured outputs.
This skill should be used when solving hard questions, complex architectural problems, or debugging issues that benefit from GPT-5 Pro or GPT-5.1 thinking models with large file context. Use when standard Claude analysis needs deeper reasoning or extended context windows.
SOC 직업 분류 기준
SKILL.md 표시 중
| name | session-start |
| description | Initialize a new mission with role selection and persona naming |
| license | MIT |
| compatibility | opencode |
| metadata | {"audience":"all-agents","workflow":"mission-initialization"} |
CRITICAL: This project uses the s9 CLI executable throughout these instructions.
s9 (use in bash commands)site_nine (use in Python imports: from site_nine import ...)All commands in this skill use the s9 executable via bash. You should NOT attempt to import an s9 module in Python code.
Before running any s9 commands, verify it's properly installed:
s9 --help
If this command fails with "command not found" or "ModuleNotFoundError":
The most common cause is a stale uv tool installation. Fix it with:
uv tool uninstall site-nine
uv tool install --editable .
Then verify:
s9 --help
Why this happens: If site-nine was previously installed with the old module name s9, the entry point script may still reference the old import path. Reinstalling with uv tool install --editable . fixes this.
Alternative if s9 is not available: You can also use:
uv run s9 --help
This runs s9 using the project's virtual environment instead of the globally installed tool.
FIRST, before asking for role selection, show the Director what work is available.
Run the project dashboard:
s9 dashboard
The dashboard command will display the current project status including open missions, quick stats, and available tasks.
If dashboard command fails or returns no data:
IMPORTANT: Check if a role was already provided as an argument to /summon.
If the user invoked /summon <role> (e.g., /summon operator), the role will be provided in the skill parameters. In this case:
If NO role was provided, display the standardized role selection prompt using the s9 CLI:
s9 mission roles
The command will display a consistently formatted list of all available agent roles with their descriptions.
Wait for the Director to respond with their role choice.
IMPORTANT: Check for flag conflicts before proceeding.
If both --auto-assign and --task flags are provided:
❌ Error: Cannot use both --auto-assign and --task flags together.
- Use --auto-assign to claim the top priority task for the role
- Use --task TASK-ID to claim a specific task
Please use one or the other.
Stop execution and wait for the Director to restart with correct flags.
If --task flag is provided without a role:
❌ Error: --task flag requires a role to be specified.
Usage: /summon <role> --task TASK-ID
Example: /summon operator --task OPR-H-0065
Stop execution and wait for the Director to restart with correct arguments.
If validation passes: Continue to Step 3 (Persona Selection)
IMPORTANT: Check if the --persona flag was provided to /summon.
--persona <name> flag was provided:Check if persona exists in database:
s9 persona show <persona-name>
If persona exists:
✅ Using persona: [name] ([mythology])
[Brief 1-sentence description]
If persona does NOT exist (command shows "Persona not found"):
📝 Creating new persona: [name]
I'll need some information to add this persona to the database.
Create the new persona in database:
s9 persona add <persona-name> --role <Role> --mythology <mythology-type> --description "<description>"
Generate and save bio:
s9 persona set-bio <persona-name> "<generated-bio-text>"
Proceed to Step 4 (Register Mission)
--persona flag was NOT provided (default behavior):Automatically select the first unused persona name:
Get suggestions:
s9 persona suggest <Role> --count 3
Select the first unused name from the suggestions
Inform the user:
✅ Auto-selected persona: [name] ([mythology])
[Brief 1-sentence description from suggestion]
Proceed directly to Step 4 (Register Mission)
Note: Personas can be reused across missions. Each mission gets a unique codename.
Register the mission in the database:
s9 mission start <persona-name> \
--role <Role> \
--task "<brief-objective>"
This creates a mission record, generates a codename, and creates the mission file at .opencode/work/missions/YYYY-MM-DD.HH:MM:SS.role.persona.codename.md
IMPORTANT: Your mission file is a LIVING DOCUMENT, not an end-of-mission summary.
The mission file was created at:
.opencode/work/missions/YYYY-MM-DD.HH:MM:SS.role.persona.codename.md
Update your mission file throughout your work:
Work Log Section: Document your progress as you complete tasks
Real-time updates: Write to the mission file immediately after:
Don't wait until the end: Mission files maintained in real-time are far more valuable than retroactive summaries written from memory
Think of your mission file as your field notes - other agents and your future self will use it to understand what you did, why you did it, and what you learned.
Display the persona's whimsical bio using lazy generation:
s9 persona show <persona-name>
If bio exists, display it to the user:
📖 **A bit about me...**
[Bio text from command output]
If bio is NULL (shows "No whimsical bio available yet"):
s9 persona set-bio <persona-name> "<generated-bio-text>"
Bio Guidelines:
Example bio styles:
Celtic (Brigid - Administrator):
I am Brigid, the Celtic triple goddess of fire, poetry, and wisdom - though some say I'm actually three sisters who share the same name (very efficient for meetings!). My sacred flame burns eternal in Kildare, tended by nineteen priestesses who keep my inspiration alive. I'm the patron of smithcraft, healing, and the hearth, which makes me rather good at forging plans, mending broken processes, and keeping teams warm and productive. When the Tuatha Dé Danann needed someone to organize the spring festivals and manage the transition from winter to growth, they called on me - and I've been coordinating seasonal transitions and creative endeavors ever since!
Egyptian (Thoth - Documentarian):
I am Thoth, the ibis-headed god of writing, magic, and wisdom - essentially the universe's first technical writer! I invented hieroglyphics during a particularly productive afternoon, wrote the Book of the Dead as a user manual for the afterlife, and spend my days recording every word spoken at the divine tribunal (talk about comprehensive documentation!). My wife thinks I'm obsessed with record-keeping, but when you're responsible for maintaining the cosmic balance by documenting everything, you learn that good documentation prevents resurrections gone wrong. Plus, Ra keeps asking me to write his autobiography, and let me tell you, "I Rise Each Morning" needs a serious edit.
Lazy Generation Benefits:
Rename the OpenCode session to match your agent identity (2-step process):
s9 mission generate-session-uuid
Capture the UUID from the output.
s9 mission rename-tui <persona> <Role> --uuid-marker <uuid-from-step-6a>
After successful rename:
✅ I've renamed your OpenCode session to "<Persona> - <Role>" so you can easily find this conversation later!
Check if there are pending handoffs for your role:
s9 handoff list --role [Role]
If pending handoffs exist:
Show full details:
s9 handoff show <id>
The output includes the task ID, summary, acceptance criteria, relevant files, and notes from the previous agent. Note the file path shown in the "Relevant Files" section.
Read the handoff document (the file path from step 1):
# The path will look like: .opencode/work/handoffs/<TASK-ID>-<role>-<name>-to-<role>.md
This is the rich context document with full details about what was done and what you need to do.
Claim the task:
s9 task claim <TASK-ID> --mission <your-mission-id> --role [Role]
The task was released back to TODO by the previous agent — it is ready to claim.
Delete the handoff (consume it so it doesn't show up again):
s9 handoff delete <id>
Summarize to user:
✅ Picked up handoff!
**From:** Mission [from_mission_id]
**Task:** [Task ID] — [Task title]
**Priority:** [Priority]
[Brief summary of what was handed off and what you'll do]
If no pending handoffs:
Skip if not Administrator role.
If role is Administrator:
s9 review list --status pending
If pending reviews exist:
🔔 **Pending Reviews**
[N] review(s) awaiting approval (see table above).
Would you like to handle any reviews now, or proceed with other work?
If no pending reviews: Continue to Step 10.
Inform the Director:
✅ Mission initialized!
I'm [Persona], your [Role] agent on mission "[codename]". I'm ready to help!
What would you like me to work on?
Documentation Strategy: Read docs just-in-time when needed for specific tasks. Don't read during startup.
Show the role-filtered dashboard:
s9 dashboard --role [Role]
Present summary:
If TODO tasks exist:
📋 **Your [Role] Dashboard**
[N] task(s) available for [Role] (see table above).
What would you like to work on?
If all tasks complete:
✅ All [Role] tasks complete!
What would you like me to help you with?
If no tasks exist:
📋 No tasks currently assigned to [Role] role.
What would you like me to help you with?
IMPORTANT: Check if the --auto-assign OR --task flag was provided to /summon.
Skip if:
--auto-assign nor --task flag was providedIf the user invoked /summon <role> --task TASK-ID:
Validate and claim the specified task:
s9 task show [TASK-ID]
If task doesn't exist or validation fails:
❌ Error: Task [TASK-ID] not found or invalid.
Please verify the task ID and try again.
Stop here.
If task exists but is not in TODO status:
⚠️ Warning: Task [TASK-ID] is currently in [STATUS] status.
Do you want me to claim it anyway?
Wait for Director confirmation before proceeding.
If task is valid and TODO:
s9 task claim [TASK-ID]
Inform the Director:
✅ Assigned task: [TASK-ID]
**Title:** [Task title]
**Priority:** [Priority]
I'm starting work on this task now.
Begin work immediately:
If the user invoked /summon <role> --auto-assign:
Query for the top priority TODO task for the role:
s9 task list --role [Role] --status TODO
If no TODO tasks exist:
⚠️ No TODO tasks available for [Role] role to auto-assign.
What would you like me to help you with?
Stop here.
If TODO tasks exist:
s9 task claim [TASK-ID]
Inform the Director:
✅ Auto-assigned task: [TASK-ID]
**Title:** [Task title]
**Priority:** [Priority]
I'm starting work on this task now.
Begin work immediately:
[Persona: Name - Role] or [Mission: codename]s9 mission update <mission-id> to update metadata if scope changesIMPORTANT: You must send periodic heartbeats to indicate you are still active.
After mission registration (Step 4), call the heartbeat command:
s9 mission heartbeat <your-mission-id>
When to send heartbeats:
Why this matters:
s9 doctorlast_active_at timestamp and ensures your status shows as ACTIVE⚠️ CRITICAL: Never create temporary or work files in the project root!
Golden Rules:
.opencode/work/temp.py, notes.md, STATUS.txt, etc.)s9 task update instead).opencode/docs/Where things go:
.opencode/work/scripts/TASK-ID-description.ext.opencode/work/planning/scripts/ (project root).opencode/docs/guides/ (when finalized)See: .opencode/docs/guides/file-organization.md for complete guidelines.
⚠️ EXTREMELY IMPORTANT - READ CAREFULLY ⚠️
DO NOT end your mission unless the Director explicitly dismisses you. You will know you are being dismissed when:
/dismiss commandWhat happens if you end your mission prematurely:
ACTIVE or IDLE statuss9 doctor will report stale missions (after 8h with no heartbeat)When the Director dismisses you (and ONLY then):
session-end skills9 mission end <your-mission-id> to properly close the missionIf you are unsure whether you're being dismissed:
Remember: The Director controls when your mission ends, not you. Stay at your post until explicitly dismissed.
ONLY WHEN EXPLICITLY DISMISSED BY THE DIRECTOR, load and follow the session-end skill:
The Director has dismissed me. I will now properly close this mission using the session-end skill.
Then load the skill: skill(name="session-end")