How do I install the Omp Delegate skill?

Run npx skillmds@latest add akhy/omp-delegate in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Omp Delegate safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Omp Delegate?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Omp Delegate free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Omp Delegate?

akhy (@akhy) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Omp Delegate by akhy — Agent Skill

Name: Omp Delegate
Author: akhy

Omp Delegate

One-shot delegation of a task to OMP (Oh My Pi) via the `omp` CLI. Use when another agent needs to run a prompt in OMP without interactive terminal management. Triggers: "delegate to omp", "send to omp", "run in omp", "hand off to omp", "use omp-delegate".

akhy 861520b 3.2 KB Updated Sep 17, 2026

File contents

OMP-Delegate

One-shot non-interactive delegation to OMP.

Core Commands

Fire-and-forget:

omp -p "<prompt>" --no-session --mode json --thinking off

Blocking (waits for agent_end):

omp -p "<prompt>" --no-session --mode json --auto-approve 2>&1 | \
  jq -r 'select(.type == "agent_end") | .messages[] |
         select(.role == "assistant") | .content[] |
         select(.type == "text") | .text' | tail -1
echo "exit: $?"

--thinking off suppresses TUI display only; thinking tokens are still generated. Stripping thinking from the output stream saves caller-side tokens, not model tokens.

Token-saving (strips all noisy events — use for high-volume delegation):

omp -p "<prompt>" --no-session --mode json --auto-approve 2>&1 | \
  jq -r 'select(.type == "agent_end") | .messages[] |
         select(.role == "assistant") | .content[] |
         select(.type == "text") | .text' | tail -1

Multi-Shot (Retained Context)

# Call 1: create session
DIR=$(mktemp -d)
OUT=$(omp -p "<prompt A>" --session-dir "$DIR" --mode json --auto-approve 2>&1)
SID=$(echo "$OUT" | python3 -c "import sys,json; [print(json.loads(l).get('id','')) for l in sys.stdin if l.strip()]" | head -1)

# Calls 2..N: resume
omp -p "<prompt B>" --session-dir "$DIR" --resume "$SID" --mode json --auto-approve 2>&1 | \
  jq -r 'select(.type == "agent_end") | .messages[] | select(.role=="assistant") | .content[] | select(.type=="text") | .text' | tail -1

Session ID (ULID) comes from session.id in the first JSON line of Call 1.
Use one --session-dir per delegated task.
--continue creates a new session — always use --resume "$SID" to continue within the same session.

Common Flags

Flag	Notes
`-p`	Prompt text
`--no-session`	Ephemeral; no session file
`--mode json`	Line-delimited JSON on stdout
`--auto-approve`	Skip tool-approval prompts
`--cwd <path>`	Working directory
`--max-time <n>`	Max duration (e.g. `600`, `10m`)
`--model <id>`	Specific model (e.g. `opus`, `openai/gpt-5.2`)
`--smol` / `--slow`	Fast/reasoning model override
`--no-tools`	Prompt-only; no tool execution
`--no-lsp`	Disable LSP tools
`--system-prompt <text>`	Override system prompt
`--append-system-prompt @<file>`	Append file to system prompt
`--session-dir <path>`	Session storage directory
`--resume <id>`	Resume specific session by ID prefix
`--print-thoughts`	Include thinking in JSON output

Output Events

type	Meaning
`session`	Session metadata (id, cwd). First line.
`agent_end`	Session complete. Completion signal.
`error`	Task failed

Exit Codes

Code	Meaning
0	Success
1	General error
2	Usage error
126	OMP not found

Error Extraction

omp -p "<prompt>" --no-session --mode json 2>&1 | grep '"type":"error"' | jq '.'

---name: omp-delegatedescription: One-shot delegation of a task to OMP (Oh My Pi) via the `omp` CLI. Use when another agent needs to run a prompt in OMP without interactive terminal management. Triggers: "delegate to omp", "send to omp", "run in omp", "hand off to omp", "use omp-delegate".---# OMP-DelegateOne-shot non-interactive delegation to OMP.## Core Commands**Fire-and-forget:**```bashomp -p "<prompt>" --no-session --mode json --thinking off```**Blocking** (waits for `agent_end`):```bashomp -p "<prompt>" --no-session --mode json --auto-approve 2>&1 | \ jq -r 'select(.type == "agent_end") | .messages[] | select(.role == "assistant") | .content[] | select(.type == "text") | .text' | tail -1echo "exit: $?"```> `--thinking off` suppresses TUI display only; thinking tokens are still generated.> Stripping thinking from the output stream saves caller-side tokens, not model tokens.**Token-saving** (strips all noisy events — use for high-volume delegation):```bashomp -p "<prompt>" --no-session --mode json --auto-approve 2>&1 | \ jq -r 'select(.type == "agent_end") | .messages[] | select(.role == "assistant") | .content[] | select(.type == "text") | .text' | tail -1```## Multi-Shot (Retained Context)```bash# Call 1: create sessionDIR=$(mktemp -d)OUT=$(omp -p "<prompt A>" --session-dir "$DIR" --mode json --auto-approve 2>&1)SID=$(echo "$OUT" | python3 -c "import sys,json; [print(json.loads(l).get('id','')) for l in sys.stdin if l.strip()]" | head -1)# Calls 2..N: resumeomp -p "<prompt B>" --session-dir "$DIR" --resume "$SID" --mode json --auto-approve 2>&1 | \ jq -r 'select(.type == "agent_end") | .messages[] | select(.role=="assistant") | .content[] | select(.type=="text") | .text' | tail -1```- Session ID (ULID) comes from `session.id` in the first JSON line of Call 1.- Use one `--session-dir` per delegated task.- `--continue` creates a new session — always use `--resume "$SID"` to continue within the same session.## Common Flags| Flag | Notes ||------|-------|| `-p` | Prompt text || `--no-session` | Ephemeral; no session file || `--mode json` | Line-delimited JSON on stdout || `--auto-approve` | Skip tool-approval prompts || `--cwd <path>` | Working directory || `--max-time <n>` | Max duration (e.g. `600`, `10m`) || `--model <id>` | Specific model (e.g. `opus`, `openai/gpt-5.2`) || `--smol` / `--slow` | Fast/reasoning model override || `--no-tools` | Prompt-only; no tool execution || `--no-lsp` | Disable LSP tools || `--system-prompt <text>` | Override system prompt || `--append-system-prompt @<file>` | Append file to system prompt || `--session-dir <path>` | Session storage directory || `--resume <id>` | Resume specific session by ID prefix || `--print-thoughts` | Include thinking in JSON output |## Output Events| type | Meaning ||------|---------|| `session` | Session metadata (id, cwd). First line. || `agent_end` | Session complete. Completion signal. || `error` | Task failed |## Exit Codes| Code | Meaning ||------|---------|| 0 | Success || 1 | General error || 2 | Usage error || 126 | OMP not found |## Error Extraction```bashomp -p "<prompt>" --no-session --mode json 2>&1 | grep '"type":"error"' | jq '.'```

Frequently asked questions