Repository Issue Activity (beta)

#37 Creates false positives on pure text-based content

#60 Feature: PII payload sanitization as complementary layer to skill vulnerability scanning

#35 Expose SkillSpector as an MCP server so agents can scan skills as a tool

#81 Feature request: OpenRouter

#76 Anthropic provider: structured-output schema rejected (minimum/maximum on integer/number)

#66 Anthropic provider fails on all semantic analyzers: endpoint rejects JSON Schema minimum/maximum keywords

#34 Failing LLM Anthropic Calls

#4 Anthropic provider silently falls back to static — LLM analyzers 400 on minimum/maximum in tool schema

#33 Calling the skill inspector from inside a skill

#14 False Report

#47 [phantomstars] Fake engagement detected on this repository

#82 Feature Request: Support for Anthropic Claude via AWS Bedrock

#75 feat(analyzer): detect skills snooping on the agent ecosystem (other skills, MCP config, agent memory)

#72 Add pre-commit hook support and GitHub Actions CI workflow

#67 fix(meta-analyzer): LLM-confirmed findings dropped when model returns end_line

#69 OpenAI-compatible endpoints that don't strictly honor response_format crash all LLM analyzers (fenced/prose JSON)

#68 feat(analyzer): no detection for container/sandbox escape (docker.sock, /proc/self/mem, cgroup)

#64 feat(analyzer): detect anti-refusal statements (jailbreak preamble)

#62 feat(analyzer): no detection for SSRF / cloud-metadata credential theft

#61 feat(analyzer): no detection for SSRF / cloud-metadata credential theft

#59 security: LLM meta-analysis can suppress high-severity static findings via prompt injection

#58 ci: add a public GitHub Actions workflow (lint, unit tests, DCO)

#57 feature: support local agent CLIs (claude/codex) as an LLM provider without an API key

#56 docs: document the machine-readable integration contract and trust model

#55 docs: DEVELOPMENT.md lists implemented analyzers as stubs; dangling SADD references