escalate

A command that sends one narrowly defined coding task to a more capable escalation agent for review or implementation. It then applies the returned change, records the lesson learned, and runs verification.

In plain words
What is it for?
Delegating one specific decision or small fix, applying the resulting diff, saving the agent’s lesson for future work, and checking the result.
Why use it?
A main coding session can get stuck on a difficult decision or localized fix. Escalating only that bounded piece provides additional help without handing over the entire session.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/rasputinkaiser/self-improvement-plugin/escalate
Clone the repo
git clone --depth 1 https://github.com/RasputinKaiser/Self-Improvement-Plugin
Per session 18 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 231 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00018 $0.00231
Opus 5 $0.00009 $0.00115
Sonnet 5 $0.00004 $0.00046
Haiku 4.5 $0.00002 $0.00023

Measured 2d ago against content hash 553dbce133e3, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

escalate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

commands/escalate.md · 21 lines

What it actually says

Argument: $ARGUMENTS — the bounded subtask to escalate.

Dispatch the escalate agent with this single, well-scoped task: "$ARGUMENTS"

Before dispatching, confirm the task is genuinely bounded (one decision or one localized fix). If it is broad, decompose it first and escalate only the slice the main session is stuck on — escalation is a scalpel, not a session swap.

When the agent returns its DIFF: and LESSON: blocks:

  1. Apply the diff (the autonomy gate will snapshot first if it touches SIPS plugin source).
  2. Record the LESSON to Memory Fabric scoped to the touched file: python3 <mf_cli> record --tier learning --title "escalation lesson: <topic>" \ --body "$LESSON" --tags lesson,escalation,frontier --scope <touched_file> so the workhorse recalls it next time and solves this class itself.
  3. Run /verify.
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 21 lines · 18 tokens per session scan A 553dbce133e3

Subscribe to this mod's changes

escalate is a command published in the GitHub repository RasputinKaiser/Self-Improvement-Plugin (6 stars, last pushed 5d ago), licensed MIT. It adds 18 tokens to every session and 231 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.