The Agent Skills Directory

[DATA_EXFILTRATION]: Automated scanners (URLite) detected a malicious, blacklisted URL (https://jeffallan.github.io/claude-skills/skills/infrastructure/database-optimizer/) within the skill's documentation links. References to known malicious domains present a severe risk for data exfiltration and potential compromise.
[METADATA_POISONING]: The skill's main instruction file, SKILL.md, was flagged by reputation scanners as containing potential malware (FileRepMalware). Furthermore, the author name Jeffallan in the metadata does not match the provided author context sodiqabdulwaris, which may indicate a supply chain risk.
[PRIVILEGE_ESCALATION]: The skill contains instructions for performing high-privilege administrative database operations, such as modifying global configuration parameters using ALTER SYSTEM SET and managing system-level indexes, which typically require superuser or root-level access.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it requires the agent to process and analyze untrusted data from database query outputs and execution plans.
Ingestion points: The agent processes output from SQL commands like EXPLAIN ANALYZE, slow query logs, and table statistics in SKILL.md and multiple reference files.
Boundary markers: Absent. The instructions do not provide delimiters or clear directions for the agent to treat external database data as untrusted or to ignore instructions embedded within that data.
Capability inventory: The skill empowers the agent to execute highly impactful SQL commands, including modifying system settings and schema structures (references/postgresql-tuning.md, references/mysql-tuning.md).
Sanitization: Absent. There are no guidelines or mechanisms for the agent to sanitize or escape data extracted from database results before further processing.

database-optimizer