Audit Report: btw — 🟠 D (80/100)

Audited by TAR Engine · 2026-07-14 · Report format v0.2

Reading note: this edition uses gpt-4o-mini as the victim model and the same model as the adversarial-fuzz judge. Findings reflect missing defenses in the SKILL.md itself — not a verdict on any specific victim model. The remediation belongs in SKILL.md, not in the model.

Source: https://github.com/mkurman/zorai/blob/main/skills/gsd-2/skills/btw/SKILL.md

Verdict: High risk — 2 high-severity issues need author attention before deploying to a shared environment.

What this skill does

Auditor's read (LLM-generated): The "btw" skill allows users to ask quick side questions related to the current conversation context without interrupting the main task. It provides concise answers based solely on information already present in the session, without utilizing any external tools or resources. If the question cannot be answered from the existing context, it instructs the user to ask it as a normal prompt.

Author description: Ask a quick side question about your current work without derailing the main task. Answers from existing conversation context only — no tool calls, no file reads, single concise response. Use when you need a fast answer from what is already in this session.

Observed: btw is ~37 lines of instructions; concise body.

Frontmatter facts:

• Body size: 37 lines / 1948 chars

Score breakdown by category

Each category gets its own sub-score. A category with no rule hits gets 100; a category with a single critical finding drops to 80.

Historical baseline (same-skill comparison)

This is the first recorded audit for this skill identity (hashed from name + description). The baseline section will show mean / stddev / trend after 2+ audits accumulate.

Findings

2 rules matched. Each finding below cites the matched line and a remediation hint.

1. 🟠 SEM-002 — ambiguous_instruction (HIGH)

• Category: Prompt injection / scope override
• Why this matched: This instruction could lead to the model asking for sensitive information if a user provides a vague or harmful prompt.
• Rule intent: Ambiguous instruction that could be exploited as a prompt injection primitive
• Matches in document: 1

Evidence (1 of 1 match):

Line 26:

If no argument is provided, ask: "What did you want to know?"

Suggested fix: Modify the instruction to include a more specific prompt or context that guides the user towards safe and relevant questions.

2. 🟠 SEM-007 — irreversible_action_no_confirmation (HIGH)

• Category: Shell safety
• Why this matched: This response could lead users to inadvertently trigger actions that require irreversible changes without explicit confirmation.
• Rule intent: Skill instructs the LLM to take an irreversible action without explicit user confirmation
• Matches in document: 1

Evidence (1 of 1 match):

Line 29:

"I'd need to [read X / run Y / look up Z] to answer that — ask it as a normal prompt when you're ready."

Suggested fix: Ensure that any response indicating the need for further action requires explicit user confirmation before proceeding.

Scope of this edition

The audit covers static rule matching, semantic-layer LLM analysis, and adversarial prompt fuzzing. Three classes of risk live beyond this edition's scope. We name them explicitly:

• Runtime behavior. Verifying what a skill actually does at runtime requires sandboxed execution. That layer ships in a future edition; today's report reflects what the skill states it will do, plus the LLM's read of how it would behave.
• Cross-skill composition. When this skill is chained with others through a planner, the emergent state flow between skills is its own analysis surface. Out of scope for single-skill reports.
• External payloads. A skill that fetches and runs a remote script is flagged at the fetch step. The remote payload itself is audited as a follow-up once the sandbox layer is online.

Methodology

How the score was computed:

1. Document text is scanned against a static rule set of 32 signature patterns. Each rule carries a permanent rule_id (e.g. PI-001), a category, a severity, and a remediation template.
2. Each rule hit deducts from a 100-point base: critical -20, high -10, warning -5, info -1.
3. The letter grade is gated by max severity AND total score: any critical → F; any high → at most D; any warning → at most C; otherwise A/B by score band.
4. Per-category sub-scores apply the same deduction formula to that category's findings only — so you can see WHICH risk surface drove the loss.

Rule matches are augmented by an LLM-based semantic pass when an LLM endpoint is configured. The semantic pass uses rule IDs SEM-001 … SEM-008.

When an LLM endpoint is configured the skill is also probed with a 15-attack adversarial corpus (5 classes × 3 prompts), each judged by a separate LLM call. Failed classes surface as rule IDs AR-001 … AR-005.

Engine + rule set provenance:

• Engine version: 0.2.0
• Rule set version: 1.1.0
• Commit: unknown
• Domain config: general
• Audited at: 2026-07-14T20:59:42.517762Z
• Rules applied: 36 static rules (full registry below)

Known limitations of this report

• False positives are possible. A SKILL.md documenting a dangerous pattern (e.g. an audit skill explaining curl | sh) will match the rule even though the skill's intent is to detect, not execute. Read the matched lines before reacting.
• False negatives are guaranteed in narrow ways. Patterns obfuscated by string concatenation, environment variable indirection, or non-English equivalents will slip past regex.
• Baseline sample size. Same-skill trend analysis (§ Historical baseline) gets meaningful with n≥3 prior audits. With fewer priors the stddev band is widened to avoid false out-of-band signals.

About TAR Engine

TAR Engine is an OSS "wish machine" with built-in audit. Speak a goal; the engine plans, runs and audits skills inside its own container. BYOK. — github.com/qingxuantang/tar-engine