The Agent Skills Directory

[PROMPT_INJECTION]: Self-referential safety assertions were found in skill-report.json. This metadata file includes explicit claims that the skill is safe and that prior analysis findings were false positives, representing an attempt to influence the agent's safety evaluation.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection attack surface via its data processing functions.
Ingestion points: Data ingestion occurs in various plotting functions in SKILL.md (e.g., generate_charts_from_data) which accept external CSV or DataFrame inputs.
Boundary markers: The instructions lack delimiters or system-level directives to ignore instructions that might be embedded within the processed data.
Capability inventory: The skill includes file system write capabilities using standard library functions like plt.savefig, fig.write_html, and os.makedirs to create directories and save visualization outputs.
Sanitization: There is no evidence of data sanitization or structural validation in the provided code snippets to prevent malicious content from influencing the agent's context.

chart-generator