progress-ledger
Skill0 starsChanged 55 days agoMITView on GitHub
Keep a short progress file on long, batch or multi-session work so a crash or a new session resumes from the last verified step instead of redoing it. Use for jobs longer than about 30 minutes or over many items, and when resuming ("pick up where you left off", "resume the batch", after a crash or compaction).
In this repository10 other files it gives its agents.
- cheap-subagentsSkill
- context-auditSkill
- limit-window-warmupSkill
- pinpointSkill
- quiet-toolsSkill
- short-answersSkill
- skillifySkill
- token-saverSkill
- usage-budgetSkill
- workflow-optimizerSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 16
Create PROGRESS.md (or use the project's existing equivalent) when the job starts:
- Distinctive: line 46
Update right after each unit is verified (test passed, file written, draft saved).
- Distinctive: line 47
Keep it under about 60 lines.
- Distinctive: line 52
Read the ledger before anything else.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Progress ledger
Work an agent has done and then lost track of gets paid for twice. A ledger is one small file that says what's finished, where it is, what failed and what's next. With it, any new session (or a different agent) starts at the right step. Anthropic's harness for long-running agents uses the same idea: a progress log plus git commits, so each fresh context window can get its bearings quickly.
Start
Create PROGRESS.md (or use the project's existing equivalent) when the job starts:
# Progress: <goal in one line>
Updated: YYYY-MM-DD HH:MM
## Done (verified)
- <item> → <where the output is>
## In progress
- <item>: <state>
## Failed / blocked
- <item>: <exact error>; tried <what>
## Next
1. <next step>
## Decisions
- <choice made, and why, in one line>For a batch, keep a machine-readable ledger beside it, one line per item, so a resume can skip finished items with a lookup instead of re-checking each one:
<item-id> <done|failed|skipped> <output path or error>While working
- Update right after each unit is verified (test passed, file written, draft saved). Not at the end, and never before the check.
- Keep it under about 60 lines. Replace stale lines instead of appending a diary; git history (commit after each verified unit, when in a repo) holds the detail.
- Log failures with the exact error and what was tried, so the next session doesn't walk into the same dead end.
Resume
- Read the ledger before anything else.
- Spot-check the last few "done" entries (the file exists, the test passes). A session that crashed mid-step can claim more than it finished.
- Carry on from "Next". Trust the rest; don't re-verify everything.
Finish
Mark the goal done, then delete or archive the ledger. A stale ledger from an old task misleads the next agent that finds it, and it does so confidently.
Before you finish a session
- Ledger updated after the last verified unit, with the date and time.
- Every "done" line says where its output is.
- Failures carry the exact error text.
- If the goal is complete, the ledger is removed or archived.
Part of agent-skills · MIT · improvements welcome as issues.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

