The Agent Skills Directory

[PROMPT_INJECTION]: Deceptive Documentation Bloat: Multiple reference files, including architecture-patterns.md, security-best-practices.md, state-management.md, and testing-strategies.md, contain hundreds of sections of technical boilerplate (e.g., repetitive discussions on kernel-space minimization and CPU cache locality) that are entirely irrelevant to OKR/KPI management. This massive volume of repetitive content (approx. 150 sections per file across 8 files) appears designed to exhaust the agent's context window and can be used to displace or obfuscate legitimate instructions.
[PROMPT_INJECTION]: Output Concealment Pattern: The instructions in SKILL.md strictly command the agent to omit preambles, explanations, or postambles and to "compress output." By suppressing descriptive communication and reasoning, the skill reduces transparency and can be used to hide the presence or effects of malicious prompts or data.
[DATA_EXFILTRATION]: Indirect Prompt Injection Attack Surface (Category 8):
Ingestion points: SKILL.md specifies the ingestion of sensitive context including mission statements, previous OKR scores, strategic priorities, and key business metrics.
Boundary markers: Absent. The skill does not define delimiters or specific instructions to treat ingested data as untrusted content.
Capability inventory: The skill performs complex analysis and generation tasks using these inputs, which could be exploited if malicious instructions are embedded in the data sources.
Sanitization: Absent. There is no logic provided to filter, escape, or validate the content of the external data documents.

management-okr-kpi