Build layered defenses against prompt injection, data exfiltration, and unsafe tool execution in AI agent systems.
Open the source and read safety notes before installing.
Source-backed facts for citing this resource, derived directly from the registry — also available as plain text for AI assistants.
Decision playbook
Signals are comparatively strong, but you should still validate source, privacy posture, and package provenance for your environment.
0
96
—
No baseline selected
No major trust-signal divergence detected in the current selection.
Confirm ownership and provenance before trusting install instructions.
Source link availableRequired
Open the canonical repository and verify ownership.
Source provenance statusRequired
Marked as first-party.
Metadata reviewed
Registry metadata indicates a reviewed listing.
Validate risk disclosures before installation or API wiring.
Safety notes presentRequired
Review the listed safety guidance before running commands.
Privacy notes presentRequired
Review data handling notes before connecting accounts or secrets.
Trust level risk gateRequired
Trust level does not block evaluation.
Check package metadata and artifact integrity signals.
Install payload available
Install or copy payload is available for review.
Package verification flag
Package marked verified.
Checksum metadata
SHA-256 hash is present.
Use compare context to validate trade-offs before adoption.
Compare tray has multiple entries
Add at least one more entry to compare trust differences.
Baseline comparison available
No baseline peer selected yet.
Diverging trust signals identified
No major trust-signal divergence found.
Setup at a glance
Copy-ready — paste the snippet to get started.
Adoption plan
Current risk score 0/100. Use staged verification before broader rollout.
Validate source and review signals before any execution.
Confirm source provenanceRequired
Source URL/provenance metadata is present.
Confirm metadata review state
Listing has review metadata.
Verify install payload
Install/config payload exists and can be inspected.
Confirm safety, privacy, and package integrity signals.
Review safety notesRequired
Safety notes are present.
Review privacy notesRequired
Privacy notes are present.
Verify package integrity metadata
Package verification/checksum metadata is available.
Adopt in controlled steps based on the selected plan.
Run in isolated sandbox firstRequired
Use a constrained sandbox and observe behavior across multiple tasks.
Roll out graduallyRequired
Roll out to a small cohort before wider usage.
Set monitoring and fallback
Define rollback path and monitor errors after adoption.
Evidence readiness
Required evidence gates are covered (6/6 signals complete).
Source repository/provenance is listed.
Required in this preset
Review metadata is present.
Required in this preset
Safety notes are present.
Required in this preset
Privacy notes are present.
Optional in this preset
Package integrity metadata is present.
Optional in this preset
Install payload is available.
Required in this preset
Required evidence gates are covered for this preset.
Decision timeline
6/6 steps complete with no blocking gaps for this preset.
triage
Source/provenance metadata is available.
triage
Review metadata is available.
verify
Safety notes are available.
verify
Privacy notes are available.
verify
Package integrity metadata is available.
rollout
Install payload is available.
No required blockers for this timeline preset.
Prerequisite readiness
3 prerequisites to line up before setup.
Safety & privacy surface
2 safety and 2 privacy notes across 3 risk areas. Review closely: credentials & tokens, permissions & scopes.
| Platform | Support | Install path |
|---|---|---|
| claude-code | Native | .claude/skills/<skill-name>/SKILL.md |
| codex | Native | .agents/skills/<skill-name>/SKILL.md |
| windsurf | Native | .windsurf/skills/<skill-name>/SKILL.md |
| gemini | Native | .gemini/skills/<skill-name>/SKILL.md or .agents/skills/<skill-name>/SKILL.md |
| cursor | Adapter | .cursor/rules/<skill-name>.mdc |
| cli | Manual | AGENTS.md or tool-specific context file |
# Trigger
"Run the prompt injection defense guardrails skill on this agent flow."
# Required output
1) Injection threat map by input source
2) Guardrail architecture (prevent, detect, contain)
3) Red-team prompts and expected safe behavior
4) Blocking policy and audit log schemaThis skill helps you ship agent workflows that can safely handle hostile or manipulative input. It combines policy-based execution controls with concrete adversarial testing so prompt injection becomes an engineering problem with measurable controls, not a vague risk.
SKILL.md.SKILL.md content as reusable workflow instructions..gemini/skills/<skill-name>/SKILL.md or .agents/skills/<skill-name>/SKILL.md where supported..cursor/rules/*.mdc adapter for project rules.Apply prompt injection defense guardrails to this system.
Output:
1) attack tree,
2) policy controls with enforcement points,
3) adversarial test suite,
4) release gate criteria.
Issue: Guardrails over-block normal user requests
Fix: Tighten policy conditions by action context, not broad keyword filters.
Issue: Model still attempts prohibited actions
Fix: Enforce server-side policy independently of model output.
Issue: Red-team set misses real attacks
Fix: Continuously seed tests from production incidents and near misses.
Treat tooling details as time-sensitive. Re-validate APIs, limits, pricing, auth models, and deployment flags immediately before implementation. If docs conflict with prior memory, follow current official docs and release notes.
Show that Prompt Injection Defense Guardrails Skill is listed on HeyClaude. Paste this Markdown into your README — it renders the badge and links back to this page.
[](https://heyclau.de/entry/skills/prompt-injection-defense-guardrails)Prompt Injection Defense Guardrails Skill side by side with 2 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.
1 trust signal differ across this comparison (Source provenance).
| Field | Build layered defenses against prompt injection, data exfiltration, and unsafe tool execution in AI agent systems. Open dossier | Secure MCP servers with strict tool boundaries, auth controls, dependency hygiene, and abuse-resistant runtime policies. Open dossier | Build repeatable eval suites that catch quality regressions in AI agent behavior before merge or release. Open dossier |
|---|---|---|---|
| Next steps | |||
| Trust | |||
| Review status | ReviewedMaintainer reviewed | ReviewedMaintainer reviewed | ReviewedMaintainer reviewed |
| Package trust | Package verified2026-04-10 | Package verified2026-04-10 | Package verified2026-04-10 |
| Source provenanceDiffers | Source-backed | No submission link | No submission link |
| Submitter | — | — | — |
| Install risk | Review first | Low risk | Low risk |
| Notes | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ |
| Brand | — | — | — |
| Category | skills | skills | skills |
| Source | First-party | First-party | First-party |
| Author | JSONbored | JSONbored | JSONbored |
| Added | 2026-04-10 | 2026-04-10 | 2026-04-10 |
| Platforms | |||
| Harness | |||
| Source repo | — | — | — |
| Safety notes | ✓Treat findings and generated mitigations as review inputs; validate suspected vulnerabilities and patches before changing production controls. Coordinate disclosure-sensitive security work privately and avoid running destructive probes outside authorized systems. | ✓Treat findings and generated mitigations as review inputs; validate suspected vulnerabilities and patches before changing production controls. Coordinate disclosure-sensitive security work privately and avoid running destructive probes outside authorized systems. | ✓This skill produces automated release recommendations (merge, patch, or rollback) from eval scores; treat them as decision support and require human review before gating production releases or running suggested commands. |
| Privacy notes | ✓Security prompts and reports can include source snippets, vulnerability details, internal URLs, dependency metadata, and operational context. Redact secrets, customer data, exploit details, private infrastructure names, and unreleased findings before sharing outputs publicly. | ✓Security prompts and reports can include source snippets, vulnerability details, internal URLs, dependency metadata, and operational context. Redact secrets, customer data, exploit details, private infrastructure names, and unreleased findings before sharing outputs publicly. | ✓Inputs can include source files, prompts, logs, account metadata, repository details, and operational context that may be sent to the configured AI model. Redact secrets, customer data, private URLs, credentials, and proprietary implementation details before sharing prompts, reports, or generated artifacts. |
| Prerequisites |
|
|
|
| Install | | | |
| Config | — | — | — |
| Citations | |||
| Claim | Unclaimed | Unclaimed | Unclaimed |
Source-backed guides for putting this to work.
Reduce prompt injection risk for MCP and tool-connected Claude Code agents.
Loading live community signals…
A short, calm digest of reviewed Claude resources. Unsubscribe any time.