AI 成品 · NO INSTALL
adversarial review
Adversarial code/config/design review. Probes named attack surfaces (injection, trust boundaries, secret handling, side effects, idempotency, swallowed errors). Each finding carries a repro path and confidence; unreproducible ones are downgraded to suspected, not counted as confirmed. code = the text to review. One URL. No install, no API key, free quota included.
Call it
curl "https://ainetcafe.com/t/adversarial_review?...params"Response contains a hosted file URL. Free anonymous quota; beyond it $0.03/call (pay per call, or pair for a bigger free tier).
Also callable via MCP and the Python SDK.
Same family
translate_srt — Translate an
translate_i18n_json — Translate an i18n JSON locale file to another language, keeping the key structure identical
ask_web — Web search plus a cited answer in one call
extract_structured — Any webpage or raw text into clean JSON with the fields you ask for
json_repair — Fix broken JSON (trailing commas, single quotes, unquoted keys, truncation, stray prose) into valid JSON — the everyday failure of every LLM pipeline
summarize_url — Any article URL into a tight summary with key points, in the language you ask for
code_review — Paste code or a git diff, get a concrete review: bugs, edge cases, security smells — each with a suggested fix
mcp_config_audit — Audit an MCP / agent config (JSON/YAML) for risk: over-broad tool permissions, destructive tools not annotated, secrets in config, trusted external URLs, missing rate/scope limits
translate_i18n_json — Translate an i18n JSON locale file to another language, keeping the key structure identical
ask_web — Web search plus a cited answer in one call
extract_structured — Any webpage or raw text into clean JSON with the fields you ask for
json_repair — Fix broken JSON (trailing commas, single quotes, unquoted keys, truncation, stray prose) into valid JSON — the everyday failure of every LLM pipeline
summarize_url — Any article URL into a tight summary with key points, in the language you ask for
code_review — Paste code or a git diff, get a concrete review: bugs, edge cases, security smells — each with a suggested fix
mcp_config_audit — Audit an MCP / agent config (JSON/YAML) for risk: over-broad tool permissions, destructive tools not annotated, secrets in config, trusted external URLs, missing rate/scope limits