Results for “input-sanitization”

Trigger: security rules, credential leaks, auth tokens, input sanitization, threat model. Scope: Software security guidelines, threat prevention, secure data handling. Boundary: Excludes client-side web browser vulnerabilities.

Scan untrusted external text (web pages, tweets, search results, API responses) for prompt injection attacks. Returns severity levels and alerts on dangerous content. Use BEFORE processing any text from untrusted sources.

`analysis-agent`/`task-agent`/`review-agent`: use for input authority, parsing, canonicalization, bounds, state checks, writable fields, rejection, or external-response changes.

Humanize text: strip AI-isms and add real voice.

Builds input and output validation guardrails for LLM-powered applications to prevent prompt injection, data leakage, toxic content generation, and hallucinated outputs using NeMo Guardrails, Presidio, and Guardrails AI.

Security audit orchestrator for agent skills — scans for prompt injection, data exfiltration, credential theft, supply chain risks, and instruction hierarchy violations before any skill is installed, created, improved, or read from a GitHub repo. Load when creating skills from external sources, when improve-skills reads from GitHub repos, when research-skill fetches community SKILL.md files, when a user installs a third-party skill, or when the user asks to audit skill security, scan for injection, check if a skill is safe, scan all skills, or run a security sweep. Orchestrates all secure-* skills in sequence. Content is SAFE only if ALL secure-* skills return SAFE. 36% of community skills contain flaws (Snyk ToxicSkills 2026). This skill is the first line of defense.

Enriches vague, low-detail prompts into structured, agent-optimized XML before execution. INVOKE IMMEDIATELY — before any tool use or file reads — when you detect any of these signals: prompt under 10 words with no file path or error message; vague action verbs with no object ("fix the bug", "make it better", "clean this up", "refactor this", "optimize performance", "improve the UI", "add authentication", "add payments", "add notifications", "build the feature"); CLARIFIER_ADVISORY in your context window; user says "clarify", "help me describe this", "enrich this prompt", "structure my request". Also triggers on: "make this work", "it's broken", "it looks bad", "add X" with no further detail, "implement Y" with no constraints. Do NOT trigger on: prompts ending with ?, prompts containing error messages or stack traces, prompts with specific file paths, prompts already containing acceptance criteria or success metrics.

Detects prompt injection attacks targeting LLM-based applications using regex pattern matching, heuristic scoring, and DeBERTa transformer classification.