Security audit
Security checks for vulnerabilities and agentic risk
The skill appears non-malicious, but its documentation mixes a general deck-structure helper with unsupported current-meta and card-purchase advice.
Install only if you want a general TCG structure helper. Treat any current metagame, matchup-ranking, card-price, or value-retention advice as needing fresh external verification; the artifact itself does not provide live data or market sources.
The manifest frames the skill as a deck-structure advisor, but the body expands into competitive deckbuilding, matchup tuning, and financial-style purchase advice. Scope drift increases the chance that an agent will apply the skill in contexts where it lacks authority, data, or safeguards, producing misleading or overconfident outputs.
The skill claims it is descriptive-only and does not use live metagame data, but later examples instruct current-meta recommendations and market-style guidance. This mismatch can mislead users and downstream agents into over-trusting unsupported advice, causing unsafe hallucinated recommendations presented as current fact.
The quality bar says outputs must stay honest about not using live metagame data, yet the examples directly depend on live tiering, matchup resources, and current-environment analysis. This internal contradiction is dangerous because it encourages deceptive presentation of stale or fabricated information as timely guidance.
Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.
def _load_prompt_template(skill_name: str) -> str:
del skill_name
return PROMPT_TEMPLATE
def _normalize_key(key: Any) -> str:
The skill description is expanded from structural deck-planning into live metagame coaching and collection-investment guidance, which introduces unsupported domains requiring current external data and financial-style recommendations. When an agent is induced to operate outside its stated competence, it is more likely to hallucinate market conditions, give outdated competitive advice, or mislead users about investment value.
The README creates conflicting instructions: the metadata says the skill should not pretend to know the live metagame, but the example asks for current-meta competitive decklists and matchup advice. This can cause the agent to overclaim current knowledge, fabricate up-to-date recommendations, or present stale advice as authoritative, which is a scope-drift and reliability vulnerability.
Without declared permissions the skill's intent is opaque and cannot be validated.
Advising users on card purchases, value retention, and investment priority goes beyond deck-structure coaching into financial recommendation territory. Even in a hobby setting, this can cause monetary harm by encouraging unsupported speculation or implied investment advice without appropriate scope, data, or disclaimers.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
'mistakes': [
'Loading up on finishers while cutting early interaction.',
'Using premium answers on low-value threats because of panic.',
'Drawing cards without checking whether the deck still has time to cash them in.',
],
},
'combo': {
The description says to use the skill when a user wants a cleaner deck plan and also frames it as applicable to 'real-life resource planning,' which broadens the trigger beyond a specific card-game context. Because the file does not provide exclusion conditions or clear boundaries for when non-TCG planning requests should not invoke the skill, activation scope is somewhat ambiguous.
The markdown includes a Chinese name and later provides a full Chinese usage scenario and expected output, but it does not state that language should follow the user's preference. This can be read as implicitly steering output toward a specific language/locale without an explicit opt-in or choice.
The normalization regex explicitly preserves CJK characters, and later parsing/scoring logic includes Chinese field names and keywords. Because the code embeds support for a specific language/locale without any visible user choice or opt-in in this file, it may conflict with a policy requiring neutral language handling or explicit language selection.
No suspicious patterns detected.