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 Vanyangyang/cursor-bridge --skill cce-routinggit clone --depth 1 https://github.com/Vanyangyang/cursor-bridgeWrote 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/vanyangyang/cursor-bridge/cce-routing)<a href="https://agentmods.dev/skills/vanyangyang/cursor-bridge/cce-routing"><img src="https://agentmods.dev/badge/skills/vanyangyang/cursor-bridge/cce-routing/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/vanyangyang/cursor-bridge/cce-routing"><img src="https://agentmods.dev/badge/skills/vanyangyang/cursor-bridge/cce-routing.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00168 | $0.01488 |
| Opus 5.5 | $0.00067 | $0.00595 |
| Sonnet 5.5 | $0.00034 | $0.00298 |
| Haiku 4.5 | $0.00017 | $0.00149 |
Grade A, and why
cce-routing 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 1mo 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 — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CCE Routing
cursor_context_engine automatically inherits the persistent CCE model and reasoning-effort default configured through cursor_model. Do not change or reset that default unless the user explicitly asks; if selection cannot be confirmed, report the pre-send failure instead of retrying with Auto.
Route project-understanding questions to the cheapest evidence surface that can answer them without guessing.
Choose CCE for project semantics
Call cursor_context_engine when one or more of these are true:
- The implementation location is unknown and finding it locally would require guessing directories or repeating broad searches.
- The question asks where or how a project behavior is implemented, what owns a state or responsibility, or why a flow behaves as it does.
- The answer requires tracing callers and callees, producer-consumer flow, configuration and registration, interface implementations, data ownership, or relationships across modules.
- An unfamiliar indexed project needs semantic context before a safe edit or review can begin.
Prefer one CCE investigation over starting an Explore subagent merely to locate or understand project code. Let Cursor choose focused or extended depth from what it discovers.
When these semantic conditions match, make CCE the first project-discovery surface. Do not establish the answer through generic context-mode, grep, or blind local exploration before trying CCE. If Claude Code denies an initial context-mode collection call with a CCE routing message, call cursor_context_engine once instead of retrying another ctx_* tool. A failed, denied, unavailable, or NOT_FOUND CCE attempt releases this priority and allows a bounded local fallback. A pre-send workspace-confirmation error may be recoverable; handle it below before falling back.
Keep deterministic work local
Do not call CCE when any of these apply:
- The exact file, symbol, or location is already known and a direct read or exact search is sufficient.
- The needed code is already present in the current context.
- The work is a trivial single-file edit or only requires running tests, inspecting logs, checking a build, or examining Git state.
- The question concerns external libraries, APIs, current documentation, or the public web.
- The user opted out of Cursor or CCE is unavailable.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago Changed · +18 lines 6f93fe9cf611
- 1mo ago Changed b13348809af0
- 1mo ago First seen · 59 lines · 168 tokens per session scan A 1633c630581d
cce-routing is a skill published in the GitHub repository Vanyangyang/cursor-bridge (82 stars, last pushed 7d ago), licensed MIT. It adds 168 tokens to every session and 1,488 once invoked, about $0.0007 per session on Opus 5.5. 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 2026-08-30.
Other skills, from other repositories
fix-default
A default procedure for handling bug fixes, organised as a three-step scan and a CP process. The input does not explain what the three steps or CP process contain.
repair-prevention-assessment
A completion gate for fixes and self-repair work. It creates a machine-checkable assessment that separates evidence a current issue is closed from evidence that the fix will remain effective.
orchestrate-test-fix
Orchestrate phases 3 and 4 - find where the built agent is wrong or fragile (edge cases, overfitting, determinism, two-strategy comparison, one-input trace), then fix wrong rows one change at a time with a diff of the whole output. Use once the pipeline runs on the sample, and whenever output rows are wrong or weakly…
audit-performance
Audit and fix runtime performance (Core Web Vitals, load priority). Use when "slow page", "LCP/INP/CLS", "fetchpriority", "long tasks", or "bfcache". JS payload → audit-bundle-size. Instant nav → enhance-web-instant-nav. Load limits → test-load.
debug-sentry-monitor
Operate Sentry: triage and fix unresolved issues, cut noise, audit instrumentation, monitor after deploy. Use when "check Sentry", "review production errors", or "improve error tracking". One bug → workflow-fix-and-ship. Plan-only audit → plan-error-handling.
housekeep-dead-code
Remove dead code by category — one commit each, typecheck and tests between, bisect on red — then add a Knip ratchet. Use when "delete the dead code", "remove unused files", or "wire up Knip". Audit first → plan-dead-code. README/deps → workflow-housekeep.