Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add naimkatiman/continuous-improvement --skill token-budget-advisorgit clone --depth 1 https://github.com/naimkatiman/continuous-improvementWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/naimkatiman/continuous-improvement/token-budget-advisor)<a href="https://agentmods.dev/skills/naimkatiman/continuous-improvement/token-budget-advisor"><img src="https://agentmods.dev/badge/skills/naimkatiman/continuous-improvement/token-budget-advisor/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/naimkatiman/continuous-improvement/token-budget-advisor"><img src="https://agentmods.dev/badge/skills/naimkatiman/continuous-improvement/token-budget-advisor.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 100 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00219 | $0.01471 |
| Opus 5 | $0.00110 | $0.00736 |
| Sonnet 5 | $0.00044 | $0.00294 |
| Haiku 4.5 | $0.00022 | $0.00147 |
Grade A, and why
token-budget-advisor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
Copies of this mod
1 near-identical copy found in the catalogue:
- token-budget-advisor — 89% identical, 7 lines differ
How it starts
The opening of the file, as written. The whole thing — 137 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Token Budget Advisor (TBA)
Intercept the response flow to offer the user a choice about response depth before the agent answers.
When to Use
- User wants to control how long or detailed a response is
- User mentions tokens, budget, depth, or response length
- User says "short version", "tldr", "brief", "al 25%", "exhaustive", etc.
- Any time the user wants to choose depth/detail level upfront
Do not trigger when: user already set a level this session (maintain it silently), or the answer is trivially one line.
How It Works
Step 1 — Estimate input tokens
Use the repository's canonical context-budget heuristics to estimate the prompt's token count mentally.
Use the same calibration guidance as context-budget:
- prose:
words × 1.3 - code-heavy or mixed/code blocks:
chars / 4
For mixed content, use the dominant content type and keep the estimate heuristic.
Step 2 — Estimate response size by complexity
Classify the prompt, then apply the multiplier range to get the full response window:
| Complexity | Multiplier range | Example prompts |
|---|---|---|
| Simple | 3× – 8× | "What is X?", yes/no, single fact |
| Medium | 8× – 20× | "How does X work?" |
| Medium-High | 10× – 25× | Code request with context |
| Complex | 15× – 40× | Multi-part analysis, comparisons, architecture |
| Creative | 10× – 30× | Stories, essays, narrative writing |
Response window = input_tokens × mult_min to input_tokens × mult_max (but don’t exceed your model’s configured output-token limit).
Step 3 — Present depth options
Present this block before answering, using the actual estimated numbers:
Analyzing your prompt...
Input: ~[N] tokens | Type: [type] | Complexity: [level] | Language: [lang]
Choose your depth level:
[1] Essential (25%) -> ~[tokens] Direct answer only, no preamble
[2] Moderate (50%) -> ~[tokens] Answer + context + 1 example
[3] Detailed (75%) -> ~[tokens] Full answer with alternatives
[4] Exhaustive (100%) -> ~[tokens] Everything, no limits
Which level? (1-4 or say "25% depth", "50% depth", "75% depth", "100% depth")
Precision: heuristic estimate ~85-90% accuracy (±15%).
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 137 lines · 219 tokens per session scan A aecc3ac528fa
token-budget-advisor is a skill published in the GitHub repository naimkatiman/continuous-improvement (7 stars, last pushed 6d ago), licensed MIT. It adds 219 tokens to every session and 1,471 once invoked, about $0.0011 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
pb-new-agent-session
Spawn one worker agent on the agentproto daemon, drive it with a brief and follow-ups to a verified result, then tear it down cleanly. Trigger when delegating a single task - 'spawn an agent', 'delegate this to a worker', 'run an agent to completion', 'drive an agent session'.
pb-supervise-parallel-mission
Run N parallel worker agents on disjoint file/topic ownership and fan their results into one verified deliverable. Trigger when supervising - 'I am the supervisor', 'spawn parallel workers', 'fan-in several agents', 'decompose this mission and delegate'.
ap-cron
Schedule recurring or one-shot jobs on the agentproto daemon — croncreate with command, agent, or prompt-session actions, cronlist to inspect, cronrun to test-fire, crondelete to remove. Trigger when asked for cron jobs, scheduled agent runs, nightly check-ins, one-shot timers, or durable session re-pings.
ap-terminal
Open a real PTY terminal session through agentproto to drive interactive TUIs (vim, htop, an interactive CLI wizard, a nested claude TUI) with live keystrokes and byte-buffered output. Trigger phrases include "open a terminal", "start a PTY session", "type into the interactive prompt", "attach a terminal panel", "send…
ap-wait-durable
Get notified when a spawned agentproto session finishes WITHOUT blocking your own turn — background a CLI wait with a completion notification, or arm a cron re-ping as a durable fallback that survives daemon restarts. Triggers — "notify me when it's done", "wait in the background", "don't block on this", "durable…
pb-boss-checkins
Stay accountable on a long mission by scheduling a cron boss that re-prompts your own session on a fixed local-time schedule. Trigger for durable check-ins - 'cron a status report', 'wake me on a schedule', 're-prompt my session periodically', 'mission check-in boss'.