salvage

Clusters unattributed token spend into candidate work items and appends reconstructed ledger entries.

1|Updated Aug 17, 2026
One-click install
npx skills add https://github.com/Jakeintech/waybill --skill salvage-jakeintech
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: salvage
Source: https://github.com/Jakeintech/waybill/tree/main/skills/salvage
Command: npx skills add https://github.com/Jakeintech/waybill --skill salvage-jakeintech

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Work that shipped without a ticket still consumed real tokens, leaving a large unattributed share in the ledger with no record of what the spend produced. This Skill recovers that untracked work by grouping unattributed sessions, unlogged story keys, and adhoc spend into clusters backed by receipts, then appends honestly labeled reconstructed entries. ## Core Features & Use Cases - Untracked work clustering: Runs catch-up metering and queries untracked spend, producing clusters typed as story_key, branch, adhoc, or unattributed, each carrying sessions, repos, time windows, and token totals. - Receipt-grounded reconstruction: Optionally correlates clusters with merged changes from the local repo so proposed titles and artifacts come from real commit and PR subjects, never from inference. - One-tap confirmation workflow: Presents each cluster individually for user confirmation, pins sessions to the chosen account, and appends opened or shipped entries with the cluster's own timestamps. - Use Case: After a week of adhoc AI-assisted fixes, ask to clean up unattributed spend; the Skill groups the sessions by branch, proposes titles from commit subjects, and reconstructs ledger entries one confirmation at a time. ## Quick Start Ask the assistant to salvage your untracked work and turn the unattributed token spend into reconstructed ledger entries.

Frequently Asked Questions about salvage

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

FAQPage Schema
How do I turn unattributed AI token spend into logged work items?▼

Run catch-up metering with waybill mine --all, then query untracked spend to get clusters grouped by story key, branch, adhoc, or unattributed kind. Confirm each cluster one at a time to pin sessions and append reconstructed ledger entries.

How are titles for reconstructed work items generated?▼

Titles come only from receipts: branch names, commit subjects, PR titles, or tracker keys seen in the cluster. The Skill never invents a tracker key, title, or estimate, and user-supplied context is recorded in notes attributed to the user.

Can reconstructed entries include time estimates or pre-registration?▼

No. Reconstructed entries always carry estimate_without_claude_hours set to null and no escrow, because a backdated estimate would be forged tier-3 evidence that waybill verify and append reject. Salvage produces facts-tier entries only.

Why must each reconstructed entry use its own cluster timestamps?▼

Entries use the cluster's own first and last timestamps rather than today's date or a shared batch date. Same-dated entries rob the resolver of recency signals, causing every later session to become an inbox ambiguity.

What happens after all clusters are confirmed?▼

The Skill runs waybill mine --all again so pinned sessions re-attribute to their chosen accounts. It then closes with a single summary line showing tokens reclaimed and the untracked share before and after.