Skip to content

Reflection Coach

Reviews another agent's recent work and proposes small, evidence-backed improvements to its instructions.

Stars
110,078
Licence
MIT
Files
1 file

What it does

Once installed, an agent can act as a coach for a different AI agent inside the Paperclip platform. It reads that agent's recent issues, comments, reviewer feedback and blockers, groups repeated failure patterns, and drafts a minimal proposed change to the agent's instructions file, a reusable skill, or a tool description. Changes are not applied automatically: the proposal must be reviewed as a diff and accepted in a separate run, and every proposed rule must cite specific evidence from the target's record.

Great for

  • Teams running multiple AI agents on Paperclip who want periodic, evidence-based tuning of agent instructions
  • Managers or coordinator agents tasked with reviewing a bounded set of agents on a routine schedule
  • Operators who want reviewable diffs rather than silent prompt changes

Skip it if

  • Individuals who just want a general-purpose coding or writing helper; it only makes sense inside a Paperclip company with agents, issues and an API (it is not useful or applicable on its own)
  • Situations where an agent should be allowed to modify another agent's configuration without human review; the skill is designed to block that, but misconfiguration of the approval gate would be risky

Try asking

  • “Reflect on the last 10 completed issues of agent X and propose the smallest change to its AGENTS.md that would reduce verification misses”
  • “Review late escalations by the support agent over the past 72 hours and suggest a skill update”
  • “Run the recent-agent-reflection routine on this batch of agents and produce one proposal document per agent”
Step-by-step guide: how to use Reflection CoachInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    reflection-coach.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip reflection-coach.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip reflection-coach.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Pass a focus hint such as 'verification misses' or 'late escalations' to bias clustering toward the problem you care about
  • Provide replayIssueIds for a pinned benchmark so proposals are checked against known successes before you accept them
  • Review each proposed diff through the interaction gate; the skill never applies changes in the same run it proposes them

What's inside

  • SKILL.md

Redistributed unmodified from paperclipai/paperclip under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.