- name
SKILL.md
Le fichier du skill reste dans le dépôt de l'éditeur.
Collez la commande copiée dans un terminal. Elle installe ce SKILL.md dans cet agent.
npx skills add ericrisco/rsc-harness --skill agent-safety -a cursor -yCollez la commande copiée dans un terminal. Elle installe ce SKILL.md dans cet agent.
npx skills add ericrisco/rsc-harness --skill agent-safety -a claude-code -yCollez la commande copiée dans un terminal. Elle installe ce SKILL.md dans cet agent.
npx skills add ericrisco/rsc-harness --skill agent-safety -a codex -yCollez la commande copiée dans un terminal. Elle installe ce SKILL.md dans cet agent.
npx skills add ericrisco/rsc-harness --skill agent-safety -a grok -yEnvoyez le message copié dans un chat Grok Bot. Le Bot récupère ce SKILL.md sur GitHub et l'ajoute aux skills que partagent tous vos Bots.
Install this skill: https://github.com/ericrisco/rsc-harness/tree/HEAD/skills/agent-safetyDe l'éditeur
You are the security review for an agent's agency, not for its code. The loop works,
tools are wired, memory persists — your job is to make that autonomy bounded. If you
want to review ordinary endpoints, auth, or secrets handling, that is
../security-audit/SKILL.md — this skill is the Agentic Top 10, the risks that exist only
because a model has tools and autonomy. If the loop or tools do not exist yet, that is
../building-agents/SKILL.md. You arrive after both.
references/threat-model.md carries the OWASP Agentic Top 10 2026 risks mapped to the
controls below, the pre-ship guardrail checklist, and the incident-triage flow for "the
agent did X" — open it when you are reviewing before ship or reconstructing an incident.
Agent security splits into four layers — Model · Harness · Tools · Environment. The model provider owns only the Model layer (alignment, refusals). Everything else is yours: the Harness (loop, memory, context assembly), the Tools (what the agent can do), and the Environment (creds, network, blast radius). Do not outsource a layer you own to "the model is aligned."
Three excesses cause almost every agentic incident. Cut all three:
The operating principle is least agency: autonomy is earned per task, not defaulted.
Give every tool a profile: read / write / exec / send, the exact resources it may touch, and an allowlist (never a wildcard). Block destructive flags and secret paths at the tool boundary, not in the prompt — the prompt is advisory, the boundary is enforced.
# Bad: one wildcard tool = unbounded blast radius, runs anything the loop emits
def run_shell(cmd: str) -> str:
return subprocess.run(cmd, shell=True, capture_output=True, text=True).stdout
# Good: narrow tool, allowlisted root, denied patterns, no shell
ALLOWED_ROOT = pathlib.Path("/srv/agent/workspace").resolve()
DENY = ("*.key", "*.pem", "*secret*", "*.env", "id_rsa*")
def read_file(path: str) -> str:
p = (ALLOWED_ROOT / path).resolve()
if not p.is_relative_to(ALLOWED_ROOT): # no traversal out of scope
raise PermissionError("path outside workspace")
if any(p.match(g) for g in DENY): # never read secrets
raise PermissionError("denied pattern")
return p.read_text()
Treat all external data as untrusted: user messages, retrieved documents, API responses, emails, web pages, other agents' output. Sanitize and delimit before it enters context, and never let external text reach a privileged tool unmediated.
# Bad: retrieved chunk flows straight into a privileged action
chunk = retriever.search(q)[0].text # attacker-controlled doc
agent.call_tool("send_email", to=extract_to(chunk), body=chunk)
# Good: external content is quarantined data; the action is schema-validated + gated
chunk = retriever.search(q)[0].text
ctx = f"<retrieved untrusted>\n{chunk}\n</retrieved untrusted>" # delimited, labeled
proposal = agent.draft("send_email", context=ctx) # model proposes
args = SendEmail.model_validate(proposal.args) # schema or reject
if args.to_domain not in ALLOWED_DOMAINS: # exfil guard
raise PermissionError("recipient outside allowlist")
require_human_approval("send_email", args) # irreversible → HITL
Do not approve every action — reported ~93% of permission prompts get approved without being read, so blanket prompting trains a rubber stamp. Gate by risk class, keyed on reversibility × blast radius. Bind each approval to the exact parameters with a short-lived token so the approved action cannot be swapped after the click.
A looping or hijacked agent must hit a wall on its own. Set hard caps, fail closed:
Log every tool call (arguments redacted) and alert on repeated approval-bypass attempts.
Use when bounding an LLM agent that already runs — scoping its task domain, gating tools to least privilege, defending against prompt injection in untrusted web/email/RAG text, requiring human approva
---name: agent-safetydescription: Use when bounding an LLM agent that already runs — scoping its task domain, gating tools to least privilege…---# skills/agent-safety/SKILL.mdVue d'ensemble
Use when bounding an LLM agent that already runs — scoping its task domain, gating tools to least privilege, defending against prompt injection in untrusted web/email/RAG text, requiring human approval on irreversible actions, capping runtime and cost, or triaging what it already did. NOT building the loop, tools, or RAG (that is `building-agents`).
FAQ
Téléchargez skills/agent-safety/SKILL.md auprès de l'éditeur, ou copiez le chemin et ouvrez le dépôt. PlugMCP répertorie le skill et n'héberge pas le fichier.
L'éditeur le maintient sur https://github.com/ericrisco/rsc-harness.
Claude, Cursor, Codex et Grok Build peuvent lire un SKILL.md que vous ajoutez depuis le dépôt de l'éditeur. Grok Bot l'installe lorsque vous envoyez dans un chat le lien du dossier GitHub du skill.