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 agentmods add skills/crumbtraildev/crumbtrail-cli/scale-cliffnpx skills add CrumbtrailDev/crumbtrail-cli --skill scale-cliffgit clone --depth 1 https://github.com/CrumbtrailDev/crumbtrail-cliWrote 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/crumbtraildev/crumbtrail-cli/scale-cliff)<a href="https://agentmods.dev/skills/crumbtraildev/crumbtrail-cli/scale-cliff"><img src="https://agentmods.dev/badge/skills/crumbtraildev/crumbtrail-cli/scale-cliff.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00084 | $0.01526 |
| Opus 5 | $0.00042 | $0.00763 |
| Sonnet 5 | $0.00017 | $0.00305 |
| Haiku 4.5 | $0.00008 | $0.00153 |
Grade A, and why
scale-cliff 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 4d 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.
How it starts
The opening of the file, as written. The whole thing — 140 lines — stays where its author put it; the contents beside it link to each section on GitHub.
The silent hard limit and scale cliff
The code is correct for the median account and fatal for the large one. A fixed timeout, a cap on bound parameters, an identifier length limit, a row maximum, a counter that overflows, or an operation whose cost grows with a dimension nobody sized for. It does not degrade. It works, and then at some size it does not.
Symptom
The reporter says it works for them and not for a colleague, or that it worked until the account grew. A request times out only for the largest tenant. An export comes back short. A list stops at a suspiciously round count. A name is accepted and then appears truncated. A job stays in progress forever for one customer and finishes everywhere else.
The tell to look for later is that failure tracks a size rather than a time.
What Crumbtrail can see
Every request carries its timing, so a session shows exactly where duration climbed and where a
request stopped completing. Named detector signals aimed at this shape include slow_request,
latency_outlier, pending_request, result_row_loss, response_exceeded_requested_limit,
response_count_mismatch, pagination_first_page_offset, accepted_text_was_truncated,
job_did_not_complete, job_drain_left_work_deferred, n_plus_one_query and
request_reconnect_storm.
Where the database adapter is installed, the session carries the rows that changed and the query activity that changed them, which is how an operation whose cost grows per row becomes visible. Where the backend integration is installed, backend spans carry the server side timing for the same request.
It does not see server memory, processor load, connection pool occupancy, garbage collection, or a database query plan. It cannot tell you the value of the limit. It can tell you the input size at which the request stopped succeeding, which is usually enough to name the dimension and let the code reveal the constant.
One capture bound matters enough to state before you use it. A response body larger than the configured maximum is summarised rather than stored, and the summary records that reason. A short looking body is therefore not proof that the application truncated anything. Read the body summary reason first, and only call it product truncation when the reason is not a capture size bound.
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.
- 4d ago First seen · 140 lines · 84 tokens per session scan A d06aef04b23e
scale-cliff is a skill published in the GitHub repository CrumbtrailDev/crumbtrail-cli (0 stars, last pushed today), licensed MIT. It adds 84 tokens to every session and 1,526 once invoked, about $0.0004 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 2026-08-31.
Other skills, from other repositories
writing
将共享历史中的已验证事实和计算结果整理成符合受众、格式与长度约束的成稿。.
feishu
Work with Feishu or Lark bots, docs, sheets, bitables, approval flows, and OpenAPI/MCP setup without hardcoding credentials.
interview
Ask one useful structured question at a time only when material product/implementation choices are genuinely missing; remember answers and produce a brief/spec. Discoverable facts should be investigated instead of asked.
test
Detect the project’s test stack, run the narrowest useful tests, create tests when authorized, and report coverage/gaps honestly.
verify
Exercise the real app/API/CLI and collect observable evidence; tests alone do not count as end-to-end verification.
ccc
This skill should be used when code search is needed (whether explicitly requested or as part of completing a task), when indexing the codebase after changes, or when the user asks about ccc, cocoindex-code, or the codebase index. Trigger phrases include 'search the codebase', 'find code related to', 'update the…