The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data (user code and infrastructure configurations) to identify cost savings.\n
Ingestion points: User-provided content via $ARGUMENTS and file review tasks in SKILL.md.\n
Boundary markers: The instructions lack explicit delimiters or warnings for the agent to ignore embedded instructions in the analyzed code.\n
Capability inventory: The skill provides instructions for using powerful AWS CLI tools (aws s3api, aws logs, aws ec2, aws budgets) which can modify account-level configurations, and uses the sharp library for file operations.\n
Sanitization: No data sanitization or validation mechanisms are described for the input code or configurations.\n- [COMMAND_EXECUTION]: The skill documents and encourages the use of various shell commands (primarily AWS CLI) to manage cloud infrastructure, such as modifying S3 storage tiers, setting log retention policies, and creating cloud budgets. These represent powerful capabilities that could be exploited if the agent is influenced by malicious input.\n- [METADATA_POISONING]: The skill includes a non-standard risk: critical field in its YAML frontmatter, which is an unconventional use of metadata. Furthermore, the skill package contains a binary file (Explanation.docx) that is not referenced anywhere in the documentation, representing an unusual structural anomaly.

cost-reducer