On-demand semantic security review of one PR before merge — reads the diff and judges intent, rather than just pattern-matching. Blocks and asks for human review on any suspicion instead of trying to resolve it. Use when asked to security-check, vet, or judge…
Benchmarking TA-Lib — ta_bench, ta_bench_direct, ta_bench_stream and scripts/stream_ab.py, what each ratio actually compares (the six-binary build matrix), streaming vs batch, and the --shape= input corpus. Use when running a benchmark in this repo,…
Autonomous codegen quality and performance loop. Use when optimizing ta_codegen C output, investigating performance regressions, or iterating toward parity with the C reference library. Triggers on "perf iteration", "performance loop", "benchmark loop",…
Use when adding a new TA-Lib indicator to ta_codegen, modifying an existing indicator's logic or metadata, or extending the generator (parser/IR/backends) for a new C construct. Triggers on "new ta func", "add indicator", "modify indicator", "next function",…