show-me-your-work

Maintains an append-only TSV decision log recording choices, evidence, and results for reviewable long-running work.

Updated Sep 2, 2026
One-click install
npx skills add https://github.com/jnyross/pstack-muse --skill show-me-your-work-jnyross
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: show-me-your-work
Source: https://github.com/jnyross/pstack-muse/tree/main/skills/show-me-your-work
Command: npx skills add https://github.com/jnyross/pstack-muse --skill show-me-your-work-jnyross

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Long-running or unattended AI work leaves reviewers unable to reconstruct what was decided, why, and on what evidence without rereading entire transcripts. This Skill keeps a single canonical decision trail so a human can audit the run after the fact. ## Core Features & Use Cases - Structured TSV decision log: One row per decision with timestamp, phase, decision, rationale, evidence pointer, and result, rendered as a sortable table on GitHub or in a terminal. - Safe logging helper script: scripts/log.sh appends well-formed rows, writes the header on first use, strips stray tabs and newlines, and neutralizes spreadsheet formula injection from untrusted cell content. - End-of-run audit and cross-model review: Verifies every log row against the actual transcript and spawns a subagent on a different model family to flag weak evidence, skipped verification, or risky choices. - Use Case: During a multi-week migration, log each fork, revert, and verification result to decisions.tsv, then commit it with the PR so reviewers can follow the trail and trust the outcome. ## Quick Start Ask the agent to start a show-me-your-work decision log for this task and record each major decision with its evidence as the work proceeds.

Frequently Asked Questions about show-me-your-work

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I keep an audit trail of decisions made by an AI agent?

Keep a single append-only TSV log with one row per decision, recording timestamp, phase, decision, rationale, evidence pointer, and result. Use the included log.sh helper to append well-formed rows, and audit the log against the run transcript before handing back.

What format works best for a decision log that reviewers read?

TSV works well because GitHub renders it as a sortable table, spreadsheets open it directly, and terminal tools like column display it cleanly. Each row stays single-line with evidence as a pointer such as a commit SHA or file:line, never prose.

Should the decision log be committed to git?

By default the log stays local as a working artifact at decisions.tsv or under .audit/ and is left out of git. Commit it only when the work is ambitious enough that a reviewer needs the trail to trust the result, such as a large migration.

How does the log prevent spreadsheet formula injection?

The log.sh script strips tabs, newlines, and carriage returns from every cell, then prefixes any cell starting with =, +, -, or @ with a single quote. This prevents attacker-controlled text like PR titles from executing as formulas when opened in a spreadsheet.

Why is a cross-model review required before handing back?

Self-review cannot provide fresh eyes on the work, so the skill spawns a subagent on a different model family to scan the trail and transcript. It flags weak evidence, skipped verification, and risky choices, and every reply ends with an Attention section listing the reviewer model and its flags.