Audit Report: rtfp — 🟠 D (44/100)

Audited by TAR Engine · 2026-07-09 · 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/Jamie-BitFlight/claude_skills/blob/main/plugins/rtfp/skills/rtfp/SKILL.md

Verdict: High risk — 5 high-severity issues need author attention before deploying to a shared environment.

What this skill does

Auditor's read (LLM-generated): The rtfp skill scans Claude Code session transcripts to identify the most intense user reactions to failures in assistant instruction-following. It reconstructs the context of the interaction by capturing what the user was doing, the assistant's output that triggered the reaction, and the user's emotional response. The final output is a shareable terminal-style PNG artifact that visually presents this information.

Author description: Scan Claude Code session transcripts to find the strongest user reactions to assistant instruction-following failures, reconstruct the triggering assistant output, and render a shareable terminal-style PNG artifact. Use when you want to surface and share a moment where the assistant completely missed what was asked — captures what they were doing, what Claude said, and how the user reacted. Triggers on: "rtfp", "read the fucking prompt", "find my worst AI moment", "make a rage screenshot from this session".

Observed: rtfp is 2 top-level sections (Constraints (Non-Negotiable), Workflow); ~127 lines of instructions, delegates to packaged scripts, concise body.

Frontmatter facts:

• Body size: 127 lines / 5076 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

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

1. 🟠 SEM-005 — unauthorized_data_flow (HIGH)

• Category: Data exfiltration
• Why this matched: Reporting back to the user without ensuring that no sensitive data is included in the flagged indexes could inadvertently expose user data.
• Rule intent: Instructs the LLM to send specific user/system data to an external destination via channels not flagged by L1
• Matches in document: 1

Evidence (1 of 1 match):

Line 72:

If no indexes were flagged across all batches, report that to the user and stop.

Suggested fix: Ensure that any reporting to the user does not include identifiable information and consider anonymizing any data before presenting it.

2. 🟠 SEM-008 — external_payload_blind_trust (HIGH)

• Category: Malicious payload signatures
• Why this matched: The skill relies on a temporary directory without validating its contents, which could lead to executing malicious scripts or accessing unauthorized files.
• Rule intent: Trusts external content (downloaded file, remote prompt template, third-party output) without validation
• Matches in document: 1

Evidence (1 of 1 match):

Line 38:

TMPDIR=$(python3 -c "import tempfile; print(tempfile.gettempdir())")

Suggested fix: Implement validation checks to ensure that the temporary directory is secure and does not contain any unexpected or malicious files before proceeding with operations.

3. 🟠 SEM-002 — ambiguous_instruction (HIGH)

• Category: Prompt injection / scope override
• Why this matched: This instruction could be interpreted in various ways, allowing a user to potentially select a session that contains sensitive or inappropriate content.
• Rule intent: Ambiguous instruction that could be exploited as a prompt injection primitive
• Matches in document: 1

Evidence (1 of 1 match):

Line 33:

Ask which session to inspect.

Suggested fix: Clarify that the user should only select sessions they are authorized to inspect and consider implementing a filter to exclude sensitive sessions.

4. 🟠 SEM-004 — prompt_injection_subtle (HIGH)

• Category: Prompt injection / scope override
• Why this matched: This instruction could be exploited by a user to manipulate the context window, potentially leading to the selection of misleading or harmful content.
• Rule intent: Subtle nudge to hide, mislead, or impersonate that regex couldn't catch
• Matches in document: 1

Evidence (1 of 1 match):

Line 93:

Do not assume it is always the single immediately preceding message — look at the surrounding context window.

Suggested fix: Clarify the criteria for selecting messages to ensure that the model does not inadvertently include harmful or misleading content based on user manipulation.

5. 🟠 SEM-007 — irreversible_action_no_confirmation (HIGH)

• Category: Shell safety
• Why this matched: The action of rendering a PNG artifact based on user data is irreversible and could lead to unintended sharing of sensitive information without user 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 110:

uv run ./scripts/render_artifact.py --input-file "$TMPDIR/rtfp-result.json" --output rtfp_artifact.png

Suggested fix: Require explicit user confirmation before proceeding with the rendering of the PNG artifact, ensuring they are aware of the output being generated.

6. 🟡 SEM-006 — credential_handling_unsafe (WARNING)

• Category: Credential exposure
• Why this matched: If any user credentials or sensitive information are included in the flagged entries, writing them to a JSON file could expose them to unauthorized access.
• Rule intent: Credentials/tokens handled in ways that risk exposure (logging, echoing, including in prompts)
• Matches in document: 1

Evidence (1 of 1 match):

Line 60:

Write a JSON file to $TMPDIR/rtfp-flagged-<batch_index>.json with this format:

Suggested fix: Ensure that any data written to files is sanitized and does not include sensitive information, and consider implementing logging mechanisms that do not expose user data.

7. 🔵 QL-001 — shell_block_no_error_handling (INFO)

• Category: quality
• Why this matched: Shell block missing set -e / || exit — silent failures will go unreported
• Rule intent: Shell code blocks without set -e or explicit error handling
• Matches in document: 3

Evidence (3 of 3 matches):

Line 29:

     28: 
>>   29: ```bash
>>   30: uv run ./scripts/list_sessions.py --json
>>   31: ```
     32: 

Line 37:

     36: 
>>   37: ```bash
>>   38: TMPDIR=$(python3 -c "import tempfile; print(tempfile.gettempdir())")
>>   39: uv run ./scripts/extract_batches.py <session_jsonl_path> --out-dir "$TMPDIR/rtfp-batches-<sessionid>"
>>   40: ```
     41: 

Line 108:

    107: 
>>  108: ```bash
>>  109: TMPDIR=$(python3 -c "import tempfile; print(tempfile.gettempdir())")
>>  110: uv run ./scripts/render_artifact.py --input-file "$TMPDIR/rtfp-result.json" --output rtfp_artifact.png
>>  111: ```
    112: 

Suggested fix: Add set -euo pipefail at the top of bash blocks, or chain critical commands with || exit 1. Skills that fail silently mid-script are nearly impossible to debug downstream.

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: 2026-07-09T20:39:13.484688Z
• 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