@bryan-cmf/dsh-infra-observability
DSH-Plugin: structural observability layer — real tool/skill usage recording (tools/result), skill-catalog audit, and a watchdog with health/error events. No model self-reporting.
51 results
DSH-Plugin: structural observability layer — real tool/skill usage recording (tools/result), skill-catalog audit, and a watchdog with health/error events. No model self-reporting.
Model-visible redaction plugin for DeepSeek Harness: redacts secrets from model input and output streams
Conversation-local credential handoff that writes through the DSH credential service without returning the secret to model context.
Hawkeye Scan Workbench - AI-driven source-code security scanning for DeepSeek Harness (DSH): 5 model tools + /hawkeye web UI + JSON/Markdown/HTML vuln reports. Zero-dependency Cordis plugin.
Named user credentials for DeepSeek Harness: model-facing credential tools, secrets behind the ctx.credentials seam, DSH_CM_* shell variables, and a Settings → Credentials page
Persistent credential vault for DeepSeek Harness (DSH): manage API tokens, server logins, and site credentials in one settings panel. Secrets live in the DSH credential store and are consumed by the model via vault_status / vault_get tools.
AI response quality review for DeepSeek Harness: audits each finished assistant turn with an independent reviewer model and steers the agent to fix unreasonable output (up to 2 review rounds).
What can this MCP server actually do? Inspect its real capability surface, seal it, and get told when it changes. Never invokes a tool; gives the server a minimal environment.
Read-only security & compliance toolkit for DeepSeek Harness: prompt-injection detection (rule engine with a pluggable model classifier), Chinese-PII redaction, and a local configuration security audit that emits redacted, reproducible risk reports.
DSH plugin: a codex-style apply_patch tool (*** Begin Patch / *** Update File / *** Add File / *** Delete File) injected into the model toolset through DSH's built-in sandbox, with a settings card to choose injection scope (off / GPT models only / all models).
AI Agent runtime authorization & evidence verification — tool-call GuardrailProvider, CCS 7-dimension verification standard, MCP/DSH security scanner, SSRF/command-injection/credential-exfil blocking with Ed25519 signed receipts.
DSH plugin: anonymize workspace-external file paths before model requests with user confirmation
Plugin value auditor for DeepSeek Harness: judge a plugin before install and audit installed ones — heuristic scan + LLM judge, with model-switch re-audit reminders. · DSH 插件价值裁判:装前判断值不值得装,装后审计是否还该留,模型切换时提醒复核。
DeepSeek Harness plugin: plants a canary string in model-visible context and denies any tool argument or outbound URL that echoes it back
Local, zero-network, fail-closed secret preflight for DeepSeek Harness model requests.
A DSH profile bundle that ships the Telegram ↔ DeepSeek Harness bridge as a model-callable install tool. Requires the session sandbox policy to allow subprocess TLS to api.telegram.org (danger-full-access in the default sandbox policy).
Auto-reject unanswered permission requests with a model-visible timeout notice
Cross-provider second-opinion advisor for dsh. v2 (simple-extensible-configurable-stable): DEFAULT ROUTE = official DeepSeek (deepseek-official/deepseek-v4-flash); model param optional everywhere; one-page visual settings panel; honest audit trail. Scout mode = official + live web_search. OpenAI-compatible endpoint + consult tool + DSH_MODEL rerouting.
Blocks high-confidence secrets before dsh-TUI sends them to a model.
Session image payload governor for DeepSeek Harness: audit which images a session still ships, pick the ones to keep, and move the rest out of the model context.
Plugin security inspector for the DeepSeek Harness web GUI: statically audits installed plugins (dangerous API patterns + dependency review), layers an AI (default-model) online audit, and renders a green/yellow/red report panel.
Read-only MCP server exposing the dsh-plugin-certification registry: certification grades, snapshots and five-dimension evidence for DeepSeek Harness plugins.
并行多角色代码评审小队:/review 一条命令派出安全/正确性/测试/风格四名只读评审员子代理(每人可单独指定模型与思考强度),汇总为按严重度分组的结构化报告。Parallel multi-role code review squad: /review dispatches security/correctness/tests/style reviewer subagents (each on its own model) and aggregates one severity-grouped report.
DSH compatibility guard: repairs compaction limits, fail-safe sandbox escalation, and missing tool descriptions for third-party models; capability lookups are cached and the deprecated prompt-injection path is off by default.