The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions define clear, task-oriented steps for producing meeting briefs and do not contain patterns intended to override agent behavior, bypass safety guidelines, or extract system prompts.- [DATA_EXFILTRATION]: The skill accesses local project artifacts (prior recaps and stakeholder summaries) to provide context for the meeting brief. There are no network operations to non-whitelisted domains or patterns suggesting the unauthorized transfer of sensitive information.- [REMOTE_CODE_EXECUTION]: No external packages or remote scripts are downloaded or executed. The skill relies on structured template rendering within the agent's safe environment.- [COMMAND_EXECUTION]: There is no evidence of subprocess spawning, shell command execution, or privileged operations like sudo. The logic is restricted to document generation.- [CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or private secrets were found in the skill or its associated templates and examples.- [OBFUSCATION]: Analysis of the skill body and referenced files revealed no hidden content, Base64-encoded strings, or homoglyph-based deception techniques.- [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests prior meeting recaps and stakeholder data. However, the capabilities are limited to markdown generation for user review, and the skill lacks dangerous tool access (network or shell), maintaining a safe profile. Evidence chain: (1) Ingestion points: Reads prior recaps and stakeholder summaries identified via project/topics frontmatter in SKILL.md. (2) Boundary markers: Absent. (3) Capability inventory: Lacks network or shell access. (4) Sanitization: Absent.

foundation-meeting-brief