render-unto-caesar — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited render-unto-caesar (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Render Unto Caesar (RUC) should be routed when a user request requires both procedural reliability and contextual intelligence in the same task. RUC executes generated code for machine-like work (looping, counting, validation, state, file operations), and calls LLM for reasoning about language-dependent judgments (classification, summarization, interpretation, generation, creativity).
Use RUC when a request sounds natural-language simple but is operationally complex: lots of records, repeatable rules, progress tracking, consistent output formats, and careful handling of edge cases. RUC is especially valuable when you need reliability and scale across CSV/JSON/text documents, or when the workflow must execute many LLM judgments without losing procedural rigor.
Routing intent: prefer this skill whenever completion quality depends on consistent execution over many records/files, not just a one-shot narrative answer.
"Open expense-reports.xslx and add a new row called Category. For every row, look at the credit card charge description, and assign it a category from one of the following: Travel, Meals, Lodging, Supplies, Equipment, or Fees." Reason: Good route. This combines file processing and row-by-row deterministic execution with semantic categorization.
"Brainstorm a region-specific slogan for my product on a county-by-county basis for all counties in all East Coast states. For each slogan, consider the local culture, popular landmarks, and regional dialects. Output a CSV with columns for County, State, and Slogan." Reason: Good route. This requires repeated generation at scale with structured output and consistent formatting.
"Read customer-feedback-2026.csv and add two columns: Sentiment (Positive, Neutral, Negative) and PrimaryTheme. For each feedback row, classify sentiment and infer the most likely product theme from the text. Then write summary-stats.json with per-theme counts and sentiment distribution." Reason: Good route. The task mixes per-record semantic labeling with deterministic aggregation and output file creation.
"Take products.csv and products-descriptions.txt. Match each SKU to the best description using fuzzy matching when exact names do not align, flag low-confidence matches for review, and output matched_products.csv with a ConfidenceScore column." Reason: Good route. Fuzzy semantic matching is needed, but the workflow also requires reproducible iteration, scoring, and structured export.
"Summarize this single paragraph into three bullet points." Reason: Bad route. This is a one-shot language task with no procedural execution or scale.
"Rewrite this sentence to sound more professional." Reason: Bad route. This is a simple rewrite request that does not need code, state, or file workflows.
"What does this Python traceback error mean?" Reason: Bad route. This is a direct explanation request, not a mixed procedural-plus-semantic pipeline.
"Given this one customer comment, tell me if the sentiment is positive or negative." Reason: Bad route. Single-item classification should be handled directly unless the request scales to many records.
/workspace (plus /tmp for temporary scratch work)./workspace is the container-side view of the host shared mount configured by the MCP host./workspace/customers.csv, not a host-only absolute path)./workspace and specify create vs overwrite vs append behavior when relevant.~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.