dsh-defend
Prompt injection, jailbreak, and secret leak defense for DeepSeek Harness with allow/ask/block interception.
Install
$ dsh plugin --profile <name> add dsh-defendPlugin Overview & Capabilities
AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.
Key Capabilities
- Aho-Corasick pattern matching and heuristic detection for prompt injections and jailbreaks
- Secret-leak detection with Shannon entropy thresholds and immediate blocking for critical secrets
- Three-way allow, ask, or block decision handling across agent/pre-step, tools/pre-execute, and tools/post-execute hooks
- Destructive recursive delete command interception for shell execution tools
- Inspection commands and reporting via defend_report tool and /defend command
- Sanitized audit logging without persisting matched sensitive text
Useful For
- Preventing prompt injection attacks from reaching the agent model via inbound messages
- Blocking accidental leakages of credentials and API keys in tool arguments or model outputs
- Guarding against unverified recursive file deletion commands in terminal execution tools
- Auditing security-relevant interception decisions within DeepSeek Harness sessions
Who It Fits
- Developers running autonomous DeepSeek Harness agents with shell or tool access
- Platform operators requiring deterministic prompt-injection and credential-leak protections
- Security teams configuring policy guardrails and audit logging for agent workflows
Documented Limitations
- Maintenance status is frozen as of 2026-10-05; no new feature development is planned
- Relies on deterministic heuristics and ported rule sets rather than semantic LLM verdicts, meaning novel attack phrasing can evade detection
- Inbound message rejection at agent/pre-step is silent and passes no error reason to the model
- Session audit event recording may be disabled by default on host harness versions lacking the ignorable event marker to prevent unresumable sessions
DSH Compatibility
Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.
Security Signals
Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.
package.json declares 0 runtime, 18 development, 7 peer, and 0 optional dependencies.
package.json declares DSH bundle metadata.
GitHub reports the repository license as Apache-2.0.
A root package.json was captured and can be inspected by the registry.
package.json declares a prepare lifecycle script that may run during relevant package installation workflows.
Public GitHub source metadata is available for this registry snapshot.
Source & Registry Notes
Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.
- Source repository
- PerryLink/dsh-defend
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- f0bbb77 · 2026-10-22
This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.
Repository Activity
- GitHub stars
- 23GitHub stars
- Forks
- 1
- Open issues
- 5
- Last commit
- 2026-10-19
- Last release
- 2026-10-18
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
Second-model AI auto-review for DeepSeek Harness approval requests with fail-closed security.
View plugin →Pre-execution safety net blocking destructive shell commands and secret leaks in coding agents.
View plugin →Persistent multi-model agent teams with dynamic DAG orchestration, run monitoring, and token insights for DeepSeek Harness.
View plugin →Native SQLite-backed task management and project board bundle with Web UI and Agent workflow tools for DeepSeek Harness.
View plugin →