usage-watcher
/usage-watcher
Reduce paid credit spend and rate-limit risk with deterministic usage analysis and budget guardrails. Use when planning high-volume agent work, reviewing recent burn, or setting lean/standard/surge operating caps.
- Author
- grtninja ↗
- Repository
- skill-arbiter ↗
- Licence
- MIT
- Last repository change
- 28.08.2026
Assessment and security
- Files in this package
- 5
- Scripts
- 1
- HTML comments
- 0
- URLs in SKILL.md
- 2
No signals detected. Repository changed on 28.08.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
- 1 scripts in this package. Review them before installing, as Claude may execute them.
Scripts: scripts/usage_guard.py
What it does and when to use it
Reduce paid credit spend and rate-limit risk with deterministic usage analysis and budget guardrails. Use when planning high-volume agent work, reviewing recent burn, or setting lean/standard/surge operating caps.
Original author description
Reduce paid credit spend and rate-limit risk with deterministic usage analysis and budget guardrails. Use when planning high-volume agent work, reviewing recent burn, or setting lean/standard/surge operating caps.
How it works
- Workflow
- Analyze Command
- Budget Plan Command
- Mandatory Chain Gate
- Mode Selection Rubric
- Guardrail Policy
- Scope Boundary
- References
- Loopback
SKILL.md preview
--- name: usage-watcher description: Reduce paid credit spend and rate-limit risk with deterministic usage analysis and budget guardrails. Use when planning high-volume agent work, reviewing recent burn, or setting lean/standard/surge operating caps. --- # Usage Watcher Use this skill to control usage cost and avoid rate-limit surprises. ## Workflow 1. Capture recent usage history to CSV/JSON. 2. Ingest live local-compute accounting from the running stack (`/health` plus `/api/accounting/summary`) whenever loopback evidence is available. 3. Run `usage_guard.py analyze` to measure burn rate, rate-limit posture, and local displacement value. 4. Run `usage_guard.py plan` to set practical daily/session caps. 5. Select and record a chain usage mode (`economy`, `standard`, `surge`) from the analysis and plan outputs. 6. Apply the recommendations before large agent workflows. ## Analyze Command ```bash python3 "$CODEX_HOME/skills/usage-watcher/scripts/usage_guard.py" analyze \ --input /path/to/usage.csv \ --window-days 30 \ --daily-budget 140 \ --weekly-budget 900 \ --credits-remaining 236 \ --five-hour-limit-remaining 100 \ --weekly-limit-remaining 0 \ --stack-health-url http://127.0.0.1:9000/health \ --stack-summary-url http://127.0.0.1:9000/api/accounting/summary \ --json-out /tmp/usage-analysis.json \ --format table ``` ## Budget Plan Command ```bash python3 "$CODEX_HOME/skills/usage-watcher/scripts/usage_guard.py" plan \ --monthly-budget 2800 \ --reserve-percent 20 \ --work-days-per-week 5 \ --sessions-per-day 3 \ --burst-multiplier 1.5 \ --json-out /tmp/usage-plan.json \ --format table ``` ## Mandatory Chain Gate Before finalizing skill chains, provide: - `usage_analysis_json=/tmp/usage-analysis.json` - `usage_plan_json=/tmp/usage-plan.json` - `usage_mode=<economy|standard|surge>` - `stack_evidence` when a local loopback stack is available If this evidence is missing, chain selection is incomplete and must fail closed. ## Mode Selection Rubric Use deterministic mode selection from analysis output: 1. `economy`: - daily status is `red`, or - weekly status is `red`, or - projected 30-day burn exceeds remaining budget. 2. `standard`: - daily/weekly status are `green` or mixed `green/yellow`, - no active governor `throttle`/`disable` recommendation. 3. `surge`: - explicit deadline pressure is present, - budget remains within guardrails, - governor action is not `throttle`/`disable`. Always record rationale in one short line with the selected mode. ## Guardrail Policy - Economy mode for discovery and triage. - Standard mode for normal implementation tasks. - Surge mode only for urgent deadlines. - Prefer bounded scripts and cached artifacts over repeated broad discovery. - When preview displacement value is positive, prefer local-first routing and skill chains that preserve paid credits and included-usage headroom. ## Scope Boundary Use this skill only for the `usage-watcher` lane and workflow defined in this file and its references. Do not use this skill for unrelated lanes; route those through `$skill-hub` and the most specific matching skill. ## References - `references/cost-control-playbook.md` - `references/usage-csv-template.csv` ## Loopback If this lane is unresolved, blocked, or ambiguous: 1. Capture current evidence and failure context. 2. Route back through `$skill-hub` for chain recalculation. 3. Resume only after the updated chain returns a deterministic next step.
Files in this package (5)
- SKILL.md 3 KB
- agents/openai.yaml 1 KB
- references/cost-control-playbook.md 2 KB
- references/usage-csv-template.csv 1 KB
- scripts/usage_guard.py 21 KB
Versions
The database has no other skills named usage-watcher.
Database snapshot: 13.11.2026 · package version f3bd876b5a43 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100ops-credit-burn-rate-watcher
Use when monitoring token and cost consumption per tenant, user, or matter across all connected LLM providers. Tracks burn rate trends, alerts when a tenant is projected to hit their plan limit within 7 days, flags abnormal per-turn costs driven by inefficient model selection or… - 100hatch3r-cost-tracking
Tracks token usage and estimate costs for agent sessions. Use when monitoring spend, approaching budget limits, or generating cost reports. - 100skill-cost-credit-governor
Govern per-skill credit and token spend with deterministic warn/throttle/disable actions. Use when usage spikes, agent chatter, or budget overruns must be detected and contained. - 100tokenmeter
Open the Tokenmeter dashboard, a local web page that tracks token usage, cost, rate-limit windows and cache misses for every prompt in Claude Code, Codex and Copilot CLI. Use when the user asks about token usage, token spend, cost per prompt or per commit, cache hit rate, usage… - 100lindy-cost-tuning
Optimize Lindy AI costs through credit management, model selection, and agent consolidation. Use when reducing spend, analyzing credit usage patterns, or optimizing budget allocation across agents. Trigger with phrases like "lindy cost", "lindy billing", "reduce lindy spend",… - 95status
Claude Code のサブスクリプション usage limit (5 時間セッション枠・週次枠の使用率と reset 時刻) を取得する。「rate limit」「残量」「usage limit」「5時間・週次枠の残りや使用率」を確認したいときに使う