agent-tooling

Deterministic fail-closed tool-call authorization for DSH with evidence: allow/block/ask policy gate plus approval-chain deferral.

Persistent, deterministic task-state primitives for DeepSeek Harness: create, status, and evidence transitions.

Bounded action-lifecycle projection for DeepSeek Harness: /action-ledger reconstructs a human-readable action ledger from the durable session log.

DeepSeek Harness health and safety diagnostics: /doctor inspects the effective model-visible tool surface.

Localhost-first HTTP debugging tool for DSH: blocks remote hosts and sensitive request headers by default.

Deterministic workspace-change attestation for DeepSeek Harness: compares bounded snapshots and reports created, changed, and removed paths.

Deterministic, privacy-aware repository context packs for DeepSeek Harness context injection.

Dependency-installation guard for DeepSeek Harness: classifies package-manager commands as ALLOW, ASK, or BLOCK before execution.

Long-running development-server lifecycle primitives for DSH; blocks compound shell usage.

该仓库暂未提供项目说明。

Structured test-result normalization for DSH: pytest, Vitest, Jest, and Cargo summaries into a stable shape.

Version-aware documentation query with excerpt bounding for DSH; host-supplied lookup, no hard-coded web provider.

Structured stack-trace parsing for DSH: Python, Node.js, Rust, and Java error frames with bounded metadata.

Offline-first static admission scanner for DeepSeek Harness community plugins: inspect a plugin package before you trust it.

Deterministic regression tests for DSH policy decisions, evaluated without invoking the protected tool body.

Lightweight read-only repository structure and symbol map for DSH; complements dsh-context-pack.

Fail-closed secret-like payload guard for DSH tool calls: inspects model tool arguments at tools/pre-execute.