Menu
One-line description of what this fix does and why it matters.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | <skill-slug> |
| description | One-line description of what this fix does and why it matters. |
| target_harness | <harness-name> |
| target_repo | <owner>/<repo> |
| target_files | ["path/to/file.ts"] |
| target_commit | <SHA or branch as of audit> |
| estimated_savings | <e.g. "30-90% input discount on Anthropic" or "0 → 80% hit rate"> |
Which harness, which file, which lines. Cite a permalink to the relevant code at the audit commit.
What's broken in the user's current setup. Be concrete:
usage field comes back zero or wrong--- a/path/to/file.ts
+++ b/path/to/file.ts
@@
- old line
+ new line
If the fix is multi-step or conditional, list the steps.
After applying, the agent should:
<command> to exercise the code path.<specific assertion> — e.g.
usage.cache_read_input_tokens > 0 on the second turn.If the assertion fails, the fix didn't land or there's a second upstream issue. Stop and report.
Why this matters in 1-2 sentences. Link to:
Aider uses 5min TTL by default and works around long pauses with keepalive pings. Wire up the 1h TTL beta instead.
Aider's --cache-prompts is off by default. Flip it on for supported models.
Stop Cline from burning a cache breakpoint on the volatile current user message every turn.
Cline OpenAI native provider sends no prompt_cache_key. Add a stable per-task key so cached_tokens stops being zero.
Cline's system prompt includes a timestamp that may be recomputed per request, invalidating the system-prompt cache.
Continue's prompt caching is opt-in via config and off by default. Flip the default to systemAndTools.