progress-tracker
Skill3 starsChanged 2 days agoMITView on GitHub
Track loop metrics, trends, improvements. Invoke: /progress-tracker.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- wpylove666/progress-trackertheir own version, first seen May 29, 2026
- mfielding92/claude-code-experttheir own version, first seen Mar 23, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- mfielding92/claude-code-expert★ 0same name
- wpylove666/progress-tracker★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository22 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- infrastructure/CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- llms-full.txtllms.txt
- llms.txtllms.txt
- chaos-agentSkill
- ci-monitorSkill
- claude-automation-recommenderSkill
- claude-md-improverSkill
- decomposeSkill
- dep-bumpSkill
- ideateSkill
- learning-loopSkill
- local-ci-precheckSkill
- md-auditSkill
- optimize-implement-queueSkill
- perf-budgetSkill
- revert-rca-loopSkill
- sentry-triageSkill
- ship-prSkill
- site-auditSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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 97
No diff, no commit, exit 0.
- Distinctive: line 101
Run this even when the day's findings are unremarkable.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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 Tracker
Queries: metrics, patterns, actions. Daily.
Queries (7d)
gh issue list --label "audit" --state all --json number,title,state,createdAt,closedAt,labels --limit 100
gh issue list --label "ci-fix" --state all --json number,title,state,createdAt,closedAt,labels --limit 50
gh issue list --label "ready" --state open --json number,title
gh issue list --label "in-progress" --state open --json number,title
gh issue list --label "has-pr" --state open --json number,title
gh issue list --label "agent-failed" --state open --json number,title
gh issue list --label "agent-skip" --state open --json number,title,createdAt
gh pr list --state merged --json number,title,mergedAt,headRefName --limit 50
gh run list --branch main --limit 20 --json conclusion,createdAt
gh issue list --label "meta-improvement" --state all --json number,title,state
cat .claude/agent-spend/sessions.jsonl 2>/dev/null | tail -100Per-issue attribution: .claude/agent-spend/sessions.jsonl = {date,timestamp,costUsd,model,adapter,status,issueNumber?,inputTokens?,outputTokens?,numTurns?} (single spend sink owned by agent-core's recordSpend seam; NOT total Claude spend — use ccusage for ground-truth totals). It only holds mbe agent run/mbe agent eval sessions. /idea-to-prod:implement-queue workers never reach recordSpend, so their usage is subagent_tokens/cost_usd in metrics/queue-telemetry.jsonl (#5885). Read both with node -e 'import("./scripts/sensors-registry.mjs").then(({SENSORS})=>{const s=SENSORS.find(x=>x.id==="agentCost");const d=s.collect({root:process.cwd(),now:new Date()});console.log(s.format(d,"agentCost"))})'. If queue cost_usd is unknown, report the tokens and say the cost is unknown. Never report it as $0.
Metrics
| Metric | Formula | Target |
| Created (7d) | audit+ci-fix | - |
| Closed (7d) | audit+ci-fix | - |
| Closure Rate | Closed/Created | >80% |
| Time-to-Close | mean(closedAt-createdAt) | <24h |
| Agent Success | has-pr/(has-pr+failed) | >70% |
| CI Pass | success/total main | >95% |
| Queue | count ready | <5 |
| Stale | ready>7d | 0 |
| Blocked | agent-failed | 0 |
| Skipped | agent-skip | 0 |
| Daily Spend | Σ costUsd (attributed) | <$10 |
| 7d Spend | Σ costUsd (attributed) | <$50 |
| Cost/Issue | 7d/closed (attributed) | <$2 |
Analysis
- Recurring types → systemic?
- Agent failures → pattern?
- CI unstable → same job?
- Queue bottleneck?
- Cost efficiency?
Log
Append .claude/improvement-loop/log.md:
## YYYY-MM-DD
### Metrics
[Table with Closed/Created/Rate/Time/Success/Pass/Queue/Stale/Skip/Spend/Cost]
### Patterns
- [Findings]
### Recommendations
- [Actions]
### Skipped Issues
[Manual review]Query skipped: gh issue list --label "agent-skip" --state open --json number,title,createdAt --jq '.[] | " #\(.number) \(.title)"'
Summary: "N skipped after max retries"
Append only.
Persist
The append above is worthless until it is committed — cloud routines run in
ephemeral checkouts, so an uncommitted log.md dies with the checkout:
node scripts/persist-metrics.mjs --routine progress-trackerIt stages every durable path with a diff (never git add -A), commits on
a branch, and opens a PR titled chore(metrics): progress-tracker <YYYY-MM-DD>
labeled has-pr — metrics-only diffs auto-merge via the low-risk fast path.
No diff, no commit, exit 0. .claude/improvement-loop/log.md is already
declared durable: true in scripts/metrics-store.mjs, so it is covered
without enumerating anything by hand.
Run this even when the day's findings are unremarkable. This step is why
#4378 happened: until it existed, this skill only ever appended, and its
entries reached main solely because /optimize-implement-queue ran later in
the same checkout and swept log.md up in its persist. When that routine
stopped producing PRs on 2026-08-16, log.md froze on the same date — nine
days with no entry, while the queue kept merging PRs daily. A step that
depends on a different skill happening to run afterwards is not a step.
Improvement Issues
Pattern 3+ days consistent?
gh issue create --title "[Meta] <improvement>" --label "meta-improvement" \
--body "## Observation
<Data>
## Recommendation
<Change>
## Impact
<Outcome>
---
*Identified: $(date +%Y-%m-%d)*"Thresholds
| Metric | Green | Yellow | Red |
| Closure | >80% | 50-80% | <50% |
| Success | >70% | 40-70% | <40% |
| CI | >95% | 85-95% | <85% |
| Queue | <5 | 5-10 | >10 |
| Time | <24h | 24-72h | >72h |
| Daily | <$10 | $10-$20 | >$20 |
| 7d | <$50 | $50-$100 | >$100 |
| Cost | <$2 | $2-$5 | >$5 |
Actions
Pause on Failure >50% (3d)
gh issue create --title "[Alert] Queue paused — failure >50%" \
--label "meta-improvement" \
--body "Success <50%. Pause loop.
$(gh issue list --label agent-failed --state open --json number,title -q '.[] | "- #\(.number): \(.title)"')
Fix, improve descriptions, re-enable."Auto-retry Stale (3+ days)
Re-queue issues that have been agent-failed for 3+ days (excludes agent-skip,
caps at 2/run). Tested logic lives in scripts/auto-retry-stale.mjs:
node scripts/auto-retry-stale.mjs # re-queue via @mbe/gh-client's markReady + comment
node scripts/auto-retry-stale.mjs --dry-run # report selection, mutate nothingThe selection rule (selectStaleForRetry) is a pure, unit-tested function
(scripts/__tests__/auto-retry-stale.test.mjs); the GitHub mutations run via
@mbe/gh-client's label machine (markReady + label.apply, #2933). Max 2/run.
Queue Adjust
Queue >10 + success >70% → /loop 15m /idea-to-prod:implement-queue
Queue 0 for 3d → reduce audit or expand
Features
gh issue list --label "feature" --state all --json number,title,state,labels --limit 100
gh issue list --label "tracking" --state open --json number,title,body% = closed/total in body.
Reverts
git log --oneline --grep="Revert" --since="7 days ago" | wc -l3/week → broken code?
Rules
- Read-only code
- Max 2 meta/run
- Max 2 retry/run
- Append-only log
- Persist before finishing:
node scripts/persist-metrics.mjs --routine progress-tracker(an uncommitted append is lost with the checkout) - Per-issue attribution:
.claude/agent-spend/sessions.jsonl(agent-core sessions, via recordSpend) +metrics/queue-telemetry.jsonl(queue workers), both read by theagentCostsensor; ccusage = ground-truth totals - Circuit: 50% over 3+ days
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.

