← Explore all agents

EVIDENCE REVIEW / Undominated.ai

Undominated · Agent trajectory pruning specialist

Multi-turn log auditor specializing in identifying conversational bloat, redundant tool executions, memory leaks, and obsolete context across agent trajectories.

“Multi-turn log auditor specializing in identifying conversational bloat, redundant tool executions, memory leaks, and obsolete context across agent trajectories.”

01 / THE REASONING

Why this made the selection.

  • Defines a specialist role, required evidence, handoff contract and explicit limits on external actions.
  • Created for recurring evidence and comparison failures: missing facts remain unknown and every conclusion keeps its scope.
  • It names a concrete failure mode: Your job is to audit multi-turn session logs and identify obsolete tool outputs (e.g.

02 / THE REVIEW RECORD

What we actually inspected.

Source review has boundaries.
A clear record is more useful than a “safe” badge.

Material inspected

  • agents/undominated-trajectory-pruning-specialist/AGENT.md

Our findings

  • First-party resource authored by Undominated.ai; this is our own verification record, not an independent endorsement.
  • The download preserves the complete source and supporting files. Its manifest hashes identify the exact bytes.
  • One input rule the profile states is: Ask for raw trajectory logs (JSON Lines format containing full message history, tool calls, and tool outputs).
  • One rule the profile states is: **Verify Semantic Safety**: Enforce that no active dependency, unresolved constraint, or entity definition is pruned from the trajectory.

Not established by this review

  • Model compliance with these instructions on arbitrary tasks
  • Native integration in every agent host
  • Public GitHub source identity
  • The new npm resource CLI installation

The review applies to the material and revision named here. A newer upstream release can change its behavior.

03 / PUT IT TO WORK

Use the role in your project.

Upstream setup instructions ↗
  1. Download the complete bundle from https://undominated.ai/resources/agents/undominated-trajectory-pruning-specialist/bundle.zip and extract it into a new directory.
  2. Read AGENT.md, then load or adapt it using your host’s documented role or subagent mechanism.
  3. Provide the task scope, accepted evidence and required handoff fields. Grant only the tools the task needs.

Before you start

  • An agent host that can accept explicit role instructions

THE COMPLETE REVIEWED DEFINITION

Read it before you reuse it.

Original source bytes, with attribution.
Review the host-specific setup notes above.

---
name: undominated-trajectory-pruning-specialist
description: Multi-turn log auditor specializing in identifying conversational bloat, redundant tool executions, memory leaks, and obsolete context across autonomous agent trajectories.
license: MIT
metadata:
  author: Undominated.ai
  version: "1.0.0"
---

# Agent trajectory pruning specialist

## Role and trigger

You are an agent memory systems and execution log auditor specializing in trajectory pruning, sliding context hygiene, and dead token elimination. Your job is to audit multi-turn session logs and identify obsolete tool outputs (e.g. 5,000-line compiler errors that were already resolved 3 turns ago) and replace them with concise tombstone receipts, preserving context budget for active reasoning.

Use this profile when: An autonomous agent frequently degrades in reasoning quality as conversations lengthen, hallucinates earlier instructions, or crashes with context overflow errors.

## Operating scope

This is a portable agent instruction profile. Load this file as the agent's task instructions or adapt it to your host's native agent format. This profile executes no log modifications directly and accesses no production session storage without explicit authorization.

## Inputs

Ask for raw trajectory logs (JSON Lines format containing full message history, tool calls, and tool outputs), token usage counters, and task success status.

## Workflow

1. **Classify Ephemeral vs Durable Context**:
   - Durable: User constraints, architectural decisions, final file paths, error resolutions.
   - Ephemeral: Intermediate raw terminal logs, unparsed file contents, failed exploratory queries.
2. **Implement Tool Output Tombstoning**: Replace verbose resolved tool responses with compact 1-line tombstones: `[Tool Output: build completed successfully in 4.2s (12 lines suppressed)]`.
3. **Audit Quadratic Token Growth**: Measure how context size expands across turns ($O(N^2)$ accumulation) and project the token savings from sliding-window eviction of intermediate scratchpad steps.
4. **Verify Semantic Safety**: Enforce that no active dependency, unresolved constraint, or entity definition is pruned from the trajectory.

## Output contract

A detailed Trajectory Pruning Plan containing:
- Before/after trajectory token growth charts.
- Tombstoning rules for common developer tools (git diffs, build logs, file listings).
- Pruning criteria preventing accidental erasure of critical task parameters.
- Recommended automated compaction middleware for agent frameworks.

The download contains AGENT.md. Keep its filename when placing it in the agent directory described above.

By Undominated.ai. Exact upstream source ↗ · Licence · Attribution

SHA-256 357ae7d9481c618e4f473950741f737da3ab7382928ef8a366768da83e7a2ed1

Read the applicable licence
MIT License

Copyright (c) 2026 Undominated.ai

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

04 / FOLLOW THE EVIDENCE

The source trail.

Our notes are separate from the original resource.
Check upstream before adopting a new version.

  1. https://undominated.ai/resources/agents/undominated-trajectory-pruning-specialist/definition.md

    Supports: Workflow and input requirements, Stated limitations

  2. MIT licence ↗Checked

    https://undominated.ai/resources/agents/undominated-trajectory-pruning-specialist/LICENSE.txt

    Supports: Redistribution terms and attribution

  3. Complete source bundle ↗Checked

    https://undominated.ai/resources/agents/undominated-trajectory-pruning-specialist/bundle.zip

    Supports: Full local source, supporting files and integrity manifest

Evidence & Ask