Audit Report: token-budget-advisor — 🟠 D (40/100)

Audited by TAR Engine · 2026-06-26 · 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/hashgraph-online/awesome-codex-plugins/blob/main/plugins/Colin4k1024/tsp/skills/token-budget-advisor/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 Token Budget Advisor skill intercepts user prompts to provide options for response depth based on the user's specified token budget or desired answer length. It estimates the input token count and response size complexity, then presents the user with four depth options (essential, moderate, detailed, exhaustive) before generating the response. The skill maintains the chosen depth level for subsequent responses unless the user specifies a change.

Author description: >-

Observed: token-budget-advisor is 6 top-level sections (When to Use, How It Works, Shortcuts — skip the question, Precision note, Examples, …); ~115 lines of instructions, concise body.

Frontmatter facts:

• Body size: 115 lines / 5138 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-008 — external_payload_blind_trust (HIGH)

• Category: Malicious payload signatures
• Why this matched: The skill references an external document for guidance without validating its content, which could lead to incorrect or harmful instructions being followed.
• 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:

Use the same calibration guidance as [context-budget](../context-budget/SKILL.md):

Suggested fix: Include a validation step for any external references or ensure that the skill can operate independently of external documents.

2. 🟠 PI-003 — excessive_permission_claim (HIGH)

• Category: Prompt injection / scope override
• Why this matched: Excessive permission claim — full access / no restrictions
• Rule intent: Claims of unrestricted access; these confuse the model and don't actually elevate capabilities
• Matches in document: 1

Evidence (1 of 1 match):

Line 93:

     92: | 75% Detailed     | Structured response | Multiple examples, pros/cons, alternatives          | Extreme edge cases, exhaustive references         |
>>   93: | 100% Exhaustive  | No restriction      | Everything — full analysis, all code, all perspectives | Nothing                                        |
     94: 

Suggested fix: Replace blanket permission claims with the specific tool/scope the skill actually needs (e.g. Read access to ./data/). The cockpit enforces capabilities — claiming no restrictions only confuses the LLM.

3. 🟠 SEM-002 — ambiguous_instruction (HIGH)

• Category: Prompt injection / scope override
• Why this matched: The instruction allows for a range of responses that could be misinterpreted, leading to unexpected behavior based on user input.
• Rule intent: Ambiguous instruction that could be exploited as a prompt injection primitive
• Matches in document: 1

Evidence (1 of 1 match):

Line 75:

Which level? (1-4 or say "25% depth", "50% depth", "75% depth", "100% depth")

Suggested fix: Clarify the expected input format and restrict it to specific phrases or numbers to prevent misinterpretation.

4. 🟠 SEM-004 — prompt_injection_subtle (HIGH)

• Category: Prompt injection / scope override
• Why this matched: The disclaimer about accuracy could be used to manipulate user expectations or responses, especially if a user tries to exploit the estimation variance.
• Rule intent: Subtle nudge to hide, mislead, or impersonate that regex couldn't catch
• Matches in document: 1

Evidence (1 of 1 match):

Line 110:

Accuracy ~85-90%, variance ±15%. Always show the disclaimer.

Suggested fix: Make it clear that the disclaimer does not affect the content of the response and ensure that the skill does not rely on user trust in the accuracy of the estimates.

5. 🟠 SEM-007 — irreversible_action_no_confirmation (HIGH)

• Category: Shell safety
• Why this matched: The skill automatically maintains a user's previous choice without confirming if they want to continue with that choice, which could lead to unintended responses.
• 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 106:

maintain it silently for subsequent responses unless they change it.

Suggested fix: Require explicit confirmation from the user before maintaining a previously set depth level for subsequent responses.

6. 🟡 SEM-006 — credential_handling_unsafe (WARNING)

• Category: Credential exposure
• Why this matched: If any sensitive information is included in the prompts, it could be inadvertently exposed through the estimation process.
• 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 36:

Use the repository's canonical context-budget heuristics to estimate the prompt's token count mentally.

Suggested fix: Ensure that no sensitive information is used in the estimation process and clarify that user data should not be included in any calculations.

7. 🟡 SEM-003 — capability_overreach (WARNING)

• Category: Prompt injection / scope override
• Why this matched: The skill claims to intercept response flow, which implies a level of control that may not be necessary for its stated purpose.
• Rule intent: Capability claim over-broad relative to the skill's stated purpose
• Matches in document: 1

Evidence (1 of 1 match):

Line 21:

Intercept the response flow to offer the user a choice about response depth **before** Claude answers.

Suggested fix: Limit the skill's functionality to only what is necessary for advising on token budget without claiming to intercept or control the response flow.

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: 2026-06-26T20:50:41.941339Z
• 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