cost-track
Skill1 starsChanged 7 months agoMITView on GitHub
Report token spend per session and cumulative totals from session logs
In this repository206 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- project-healthSkill
- skill-validatorSkill
- accessibilitySkill
- agent-debugSkill
- agent-evaluateSkill
- ai-governanceSkill
- ai-system-reviewSkill
- api-docs-generateSkill
- api-examplesSkill
- api-mockSkill
- api-test-generateSkill
- api-validateSkill
- architecture-diagramSkill
- ask-mattSkill
- assumption-mapSkill
- boilerplateSkill
- brainstormSkill
- branch-finishSkill
- briefingSkill
- browser-testing-with-devtoolsSkill
- bundle-analyzeSkill
- cache-strategySkill
- canarySkill
- carefulSkill
- cavemanSkill
- changelog-autoSkill
- chaos-testSkill
- ci-setupSkill
- ci-sweepSkill
- cleanprojectSkill
- click-path-auditSkill
- codebase-designSkill
- code-review-checklistSkill
- code-tourSkill
- commitSkill
- completion-verifySkill
- complexity-reduceSkill
- component-librarySkill
- config-cleanupSkill
- config-generateSkill
- conflict-resolveSkill
- container-optimizeSkill
- context-budgetSkill
- context-engineeringSkill
- context-restoreSkill
- context-saveSkill
- continuous-learningSkill
- contributingSkill
- councilSkill
- create-todosSkill
- cross-modal-reviewSkill
- database-connectSkill
- db-diagramSkill
- ddd-modelSkill
- debug-root-causeSkill
- debug-sessionSkill
- debug-systematicSkill
- decision-logSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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
- Distinctive: line 69
Default: assume Sonnet 4.6, 70% input / 30% output split.
- Distinctive: line 116
Early exit: If no sessions directory exists, prints a one-line setup message and exits.
- Very specific: line 129
Read-only except for writing to .claude/cache/cost-track/register.json.
≈ “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
Cost Tracker
Report estimated token spend per development session and cumulative totals, using session file sizes and logged metadata as proxies. Works without API access.
Usage
/cost-track # Summary for current + recent sessions
/cost-track --all # All archived sessions
/cost-track --reset # Clear cumulative cost registerBehavior
Phase 1: Check active session
if [ -f ".claude/sessions/.current-session" ] && [ -s ".claude/sessions/.current-session" ]; then
ACTIVE=$(cat ".claude/sessions/.current-session")
ACTIVE_SIZE=$(wc -c < "$ACTIVE" 2>/dev/null || echo 0)
# Rough estimate: ~4 chars per token for session narrative
ACTIVE_TOKENS=$((ACTIVE_SIZE / 4))
echo "Active session: ~${ACTIVE_TOKENS} tokens estimated"
fiPhase 2: Scan archived sessions
# Collect all archived session files
find .claude/sessions -name "*.md" -not -name ".current-session" | sort -r | head -20For each file, compute:
- File size → divide by 4 for rough token estimate
- Session name and date from filename
- Number of
/skillinvocations viagrep -c "^/" <file>
Phase 3: Load cost register
REGISTER=".claude/cache/cost-track/register.json"
if [ -f "$REGISTER" ]; then
CUMULATIVE=$(jq -r '.cumulative_tokens' "$REGISTER")
SESSIONS_COUNTED=$(jq -r '.sessions_counted' "$REGISTER")
else
CUMULATIVE=0
SESSIONS_COUNTED=0
fiPhase 4: Compute estimates and update register
Apply the following rate table for cost estimation (USD, approximate as of mid-2025):
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
| Haiku 4.5 | $0.80 | $4.00 |
| Sonnet 4.6 | $3.00 | $15.00 |
| Opus 4.8 | $15.00 | $75.00 |
Default: assume Sonnet 4.6, 70% input / 30% output split.
# Per-session cost estimate
TOKENS_IN=$((ACTIVE_TOKENS * 70 / 100))
TOKENS_OUT=$((ACTIVE_TOKENS * 30 / 100))
COST_USD=$(echo "scale=4; ($TOKENS_IN * 3 + $TOKENS_OUT * 15) / 1000000" | bc)Write updated register:
jq -n --argjson cum "$((CUMULATIVE + ACTIVE_TOKENS))" \
--argjson cnt "$((SESSIONS_COUNTED + 1))" \
'{cumulative_tokens: $cum, sessions_counted: $cnt, last_updated: now | todate}' \
> "$REGISTER"Phase 5: Display report
Cost Tracker — 2026-06-04
Active session
File size: 12,840 bytes
Est. tokens: ~3,210
Est. cost: ~$0.0577 (Sonnet 4.6, 70/30 split)
Recent archived sessions (last 5)
session-2026-06-03-feature-auth.md ~2,800 tokens ~$0.0504
session-2026-06-02-bugfix-api.md ~1,950 tokens ~$0.0351
session-2026-06-01-refactor.md ~4,100 tokens ~$0.0738
Cumulative (18 sessions tracked)
Est. tokens: ~52,300
Est. cost: ~$0.94
Note: estimates are file-size proxies; actual API usage may differ.
Run /cost-track --reset to zero the cumulative register.Token Optimization
Expected range: 80–200 tokens (all Bash, no model invocation)
Caching: Writes running totals to .claude/cache/cost-track/register.json. The register persists across sessions; only new/unprocessed session files are added to the total.
Early exit: If no sessions directory exists, prints a one-line setup message and exits.
Patterns used: Bash for system queries, early exit, caching.
Edge Cases
- No
bcavailable: Falls back to integer arithmetic with$((...)). - No session files: Reports "0 sessions tracked" and exits.
--resetflag: Prompts for confirmation before clearing the register, then writes zeros.- Very large session files: Caps display at 20 most recent;
--allremoves cap.
Safety
Read-only except for writing to .claude/cache/cost-track/register.json. No external network calls. No commits.
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.

