moonlight-lupin/agent-skills / research/endpoint-probe
endpoint-probe
FlaggedUse when you need to discover an API's surface — poke a base URL to find its type (REST, GraphQL, SOAP, JSON-RPC), auth scheme, available endpoints, rate limits, and CORS policy. Also probes MCP servers (HTTP and stdio) to discover tools, resources, and prompts with full input schemas. Runs a Python probe script that checks well-known discovery paths (OpenAPI/Swagger, GraphQL introspection, health, .well-known), fingerprints the server, sends OPTIONS for CORS, detects auth from 401/403 responses, and guesses common REST resources. For MCP, performs JSON-RPC initialize handshake then tools/list, resources/list, prompts/list. Feed it a base URL or MCP server URL, optionally with auth credentials.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 18.9 KB |
| references/code-review-history.md | markdown | 2.9 KB |
| references/code-review-v1.2.1.md | markdown | 4.1 KB |
| references/interpretation-guide.md | markdown | 4.5 KB |
| scripts/probe_api.py | script | 67.6 KB |
| tests/conftest.py | script | 185 B |
| tests/test_redaction.py | script | 4.8 KB |
references/code-review-history.md
View on GitHub# Code Review History — api-discovery v1.2.3
6 rounds of iterative review→fix→re-review (2026-07-16).
## Reviewers
| Round | Reviewer | Model | Fixer |
|---|---|---|---|
| R1 | cursor-grok-4.5-high | Grok 4.5 | Orchestrator (GLM-5.2) |
| R2 | cursor-grok-4.5-high | Grok 4.5 | Orchestrator (GLM-5.2) |
| R3 | cursor-grok-4.5-high | Grok 4.5 | Orchestrator (GLM-5.2) |
| R4 | cursor-grok-4.5-high | Grok 4.5 | Codex 5.6 (gpt-5.5) |
| R5 | cursor-grok-4.5-high | Grok 4.5 | Orchestrator (GLM-5.2) |
| R6 | claude-opus-4.8 | Opus 4.8 | Orchestrator (GLM-5.2) |
## Findings by round
### R1 (4 BLOCKING, 9 MAJOR, 10 MINOR, 7 NIT)
- B1: stdio readline() no timeout — hang forever
- B2: _send re-sends on server notifications
- B3: No SSRF protection
- B4: --mcp-stdio launches arbitrary commands
- M1-M9: No Mcp-Session-Id, notification with id, no try/finally, shared header mutation, unbounded body, stdio/HTTP parse divergence, hardcoded root env, wrong TLS docs, credential exposure
### R2 (3 BLOCKING, 7 MAJOR, 10 MINOR, 7 NIT)
- B1: stderr.read() unbounded hang
- B2: IPv4-mapped IPv6 SSRF bypass
- B3: Absolute path allowlist bypass
- M1-M7: No try/finally, no id correlation, env scrub misses common secrets, DNS TOCTOU, 172.16 false positives, SKILL.md --version, stale interpretation-guide
### R3 (2 BLOCKING, 5 MAJOR, 10 MINOR, 7 NIT)
- B1: 0.0.0.0/:: unspecified not blocked
- B2: interpretation-guide still stale (claimed fixed but wasn't)
- M1-M5: -c/-e flags still RCE-capable, stdio soft-fail not applied, DNS TOCTOU undocumented, docstring --version, BrokenPipe crash
### R4 (1 BLOCKING, 3 MAJOR, 6 MINOR, 5 NIT) — fixed by Codex 5.6
- B1: except block KeyError on summary (report defaults not initialized)
- M1: stdio soft-fail for method-not-found
- M2: dangerous flag exact-token only (-cprint(1) bypasses)
- M3: version strings 1.2.2 vs 1.2.3
### R5 (1 BLOCKING, 2 MAJOR, 2 MINOR, 4 NIT)
- B1: npx --call bypasses denylist
- M1: --mcp-env NODE_OPTIONS RCE
- M2: soft-fail "not found" too broad
### R6 (1 BLOCKING, 0 MAJOR, 2 MINOR, 2 NIT) — claude-opus-4.8
- B1: Python short-flag clustering (python3 -Ic "code") bypasses all prefix checks
- Root cause: CPython _PyOS_GetOpt allows clustering no-argument flags before -c
- Fix: regex `^-[A-Za-z]*[cep]([=].*)?$` catches clustered flags
- Also added glued-code regex `^-c\S+$|^-e\S+$|^-p\S+$`
## Key patterns
1. **Denylist regression** — each denylist fix opened a new bypass (R3→R4→R5→R6). Final fix used regex instead of prefix matching.
2. **Cross-model diversity** — Opus 4.8 caught the clustering bypass that Grok missed for 5 rounds. Different model = different parser mental model.
3. **Claimed-but-not-applied fixes** — R3 B2 (interpretation-guide) and R4 M1 (stdio soft-fail) were claimed fixed but weren't. Always verify by reading the actual file.
4. **Codex sandbox** — skills-directory paths need `--sandbox danger-full-access` (bwrap mount fails with --full-auto).Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |