retro

A retrospective analysis workflow for reviewing completed multi-agent work, called a fellowship. It examines recorded events, alerts, and Git metrics, which are measurements from the version-control history.

In plain words
What is it for?
Use it after a fellowship finishes or is disbanded to review quest worktrees, phase events, alerts, and Git activity, then consider recommendations one at a time.
Why use it?
It helps identify recurring problems in how work was organized and suggests configuration improvements based on recorded evidence. It stops when no fellowship history is available.

Skill for Claude CodeCodex

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 skills/justinjdev/fellowship/retro
Any agent
npx skills add justinjdev/fellowship --skill retro
Clone the repo
git clone --depth 1 https://github.com/justinjdev/fellowship

Made for: Claude Code, Codex.

Per session 32 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,490 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 $0.00032 $0.01490
Opus 5 $0.00016 $0.00745
Sonnet 5 $0.00006 $0.00298
Haiku 4.5 $0.00003 $0.00149

Measured 2d ago against content hash ea5104a53c97, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

retro 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.

plugin/skills/retro/SKILL.md · 145 lines

How it starts

The opening of the file, as written. The whole thing — 145 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Retro — Post-Fellowship Retrospective

Overview

Analyzes a completed fellowship's history to surface patterns and interactively offer configuration improvements. Reads gate events, palantir alerts, and git metrics across all quest worktrees, then presents findings and walks through actionable recommendations one by one.

When to Use

  • Manual: invoke /retro after a fellowship completes or is disbanded
  • Suggested: Gandalf suggests running /retro during the disband flow (not enforced)

Process

Step 1: Locate Fellowship Data

  1. Find the git root directory
  2. Read ~/.claude/fellowship.json to check for a custom dataDir setting. If absent, use the default data directory (.fellowship/).
  3. Read fellowship-state.json from the resolved data directory to enumerate all quest worktrees and their metadata
  4. If no fellowship state file exists, report "No fellowship state found — nothing to analyze" and stop

Step 2: Collect Data

For each quest worktree listed in fellowship-state.json:

  1. Gate events: Read quest-herald.jsonl from the worktree's data directory via the Read tool. Each line is a JSON object with timestamp, quest, type, phase, and detail fields. Collect all entries of type gate_approved, gate_rejected, gate_submitted, and phase_transition.

  2. Quest state: Read quest-state.json from the worktree's data directory. Record the final phase, quest_name, and whether the quest completed.

  3. Quest tome: Read quest-tome.json from the worktree's data directory if it exists. Record gate history (approved/rejected counts per phase), phases completed with durations, and files touched.

  4. Git metrics: Run these Bash commands for each worktree:

    • git -C {worktree} log --oneline | wc -l — commit count
    • git -C {worktree} diff --stat "$(git -C {worktree} rev-list --max-parents=0 HEAD | tail -n1)"..HEAD 2>/dev/null || echo "0 files changed" — change summary
  5. Palantir alerts: Read palantir-alerts.jsonl from the resolved data directory at the git root if it exists. Each line is a JSON object with timestamp, type (stuck/drift/conflict/health/bulletin), and detail. Standard alerts include quests; bulletin alerts include source_quest, target_quest, topic, and discovery instead.

Read the full file on GitHub · 145 lines

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 · 145 lines · 32 tokens per session scan A ea5104a53c97

Subscribe to this mod's changes

retro is a skill published in the GitHub repository justinjdev/fellowship (5 stars, last pushed 19d ago), licensed Apache-2.0. It adds 32 tokens to every session and 1,490 once invoked, about $0.0002 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.