show-me-your-work

Maintains an append-only TSV decision log with evidence for long-running or unattended work.

Updated Sep 23, 2026
One-click install
npx skills add https://github.com/lucasheriques/pstack --skill show-me-your-work-lucasheriques
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: show-me-your-work
Source: https://github.com/lucasheriques/pstack/tree/main/skills/show-me-your-work
Command: npx skills add https://github.com/lucasheriques/pstack --skill show-me-your-work-lucasheriques

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve? Long-running, autonomous, or multi-phase engineering work leaves no reviewable record of what was decided and why, making it hard for a human reviewer to trust the result after stepping away. ## Core Features & Use Cases - Decision Log Format: Defines a single TSV file with one row per decision, capturing timestamp, phase, decision, reason, evidence pointer, and result. - Logging Helper Script: Provides scripts/log.sh to append well-formed rows with automatic timestamps, header creation, and spreadsheet formula-injection protection. - Audit and Cross-Model Review: Verifies the log against the session transcript and spawns a read-only subagent on a different model to flag weak evidence or risky choices. - Use Case: During an unattended multi-hour migration, each fork, revert, and verification result is logged with a commit SHA or file path, so a reviewer can later audit the trail row by row. ## Quick Start Ask the AI to keep a show-me-your-work decision log for this long-running task and record each major decision with its evidence.

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 a decision log for a long-running AI coding task?▼

Create a TSV file with columns ts, phase, decision, why, evidence, and result, then append one row per decision point using the log.sh helper script. Log forks, completed units with verification results, pivots, and blockers rather than every trivial action.

What format should an audit trail for autonomous work use?▼

Use a single TSV file where each row records an ISO8601 timestamp, the work phase, a one-line decision, a plain-language reason, an evidence pointer like a commit SHA or file:line, and the outcome. Cells stay single-line and evidence is a pointer, never prose.

Should the decision log be committed to git?▼

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

How is the decision log verified before handing back results?▼

The log is audited against the session transcript to confirm every row maps to a real action and each evidence pointer resolves. A read-only subagent on a different model then reviews the trail and flags weak evidence, skipped verification, or risky choices.

Why does the logging script prefix cells starting with =, +, -, or @?▼

Spreadsheet applications parse cells starting with those characters as formulas, so attacker-controlled text like PR titles could execute when a reviewer opens the file. The script prefixes such cells with a single quote to prevent formula injection.