Grades past shipped runs against git history, traces, and the pact ledger.
npx skills add https://github.com/azure-id/orc --skill orc-aftermath-azure-id
Please help me install this Agent Skill. Skill: orc-aftermath Source: https://github.com/azure-id/orc/tree/main/templates/skills/orc-aftermath Command: npx skills add https://github.com/azure-id/orc --skill orc-aftermath-azure-id
git log, ORC traces, and the pact ledger. - Evidence-backed reports: Writes orc-aftermath/<period>/aftermath.md where every signal names its files, commits, or pact ids, plus a structured block that /orc-retro aggregates. - Blame-free by design: Reports signals and their strength, never verdicts, never person names, and marks runs under 7 days old as too recent to grade. - Use Case: Ask "what did we ship last month and how did it go" and receive a per-run report comparing each run's promised acceptance criteria against what the repository now shows. ## Quick Start Ask the agent to run /orc-aftermath to grade the last 30 days of shipped runs and write the aftermath report.High-intent search queries and answers about installing and using this skill.
Run /orc-aftermath to grade past runs against the repository's own future. It inspects git log for reverts and rewritten files, checks whether added tests still exist unskipped, and reads the pact ledger for broken promises, then writes an evidence-backed report.
It uses six signals with strength ratings: reverted commits and deleted tests (strength 3), broken pact promises (3), skipped tests (2), three or more rewritten files (2), and one to two rewritten files (1, labelled weak). Every signal must carry concrete evidence.
No. It works entirely from git log, ORC's own trace corpus, and the pact ledger file. There is no vendor integration, no telemetry, and no production instrumentation required.
Any run younger than 7 days is marked too recent to grade because churn signals need time to surface. This is treated as a valid answer that keeps its slot in the report, not a gap or failure.
No. The lane is strictly read-only and report-only. It writes exactly one markdown report file and never edits code, config, rubrics, or runs, and it never names individual commit authors.
BedaBeda-GrowthDecompose plan files into Jules-sized tasks and register them in the work ledger.
ScripsteamAggregate Jira stories and git activity into a structured sprint retrospective.
manuscoRuns structured hiring workflows from outcome scorecards through interviews, offers, and 90-day onboarding.
NagyViktConfigure structured logging, tracing, and performance monitoring for Vercel-hosted applications.