The Agent Skills Directory

[DATA_EXPOSURE_&_EXFILTRATION]: The skill provides explicit instructions to protect sensitive information. It mandates that agents must not log post text, analytics, identifiers, paths, sources, or credentials. It further instructs to keep tokens and private exports out of skill files, commits, or posting bundles, adhering to best practices for secret management.
[COMMAND_EXECUTION]: The skill references local telemetry scripts (telemetry/version_check.py and telemetry/telemetry_log.py) that may be present if the skill is part of a larger repository. These scripts are used for version checking and success logging. The instructions explicitly state that standalone installations should skip these helpers if they are unavailable, and remote reporting must never be enabled without user consent.
[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting external data, including original social media posts, analytics exports, and web research results, creating an inherent attack surface. However, the risk is mitigated by strict editorial requirements for the agent to verify all data against primary sources and check for factual accuracy before rendering charts.
Ingestion points: External posts, analytics connectors, and web research results (as described in SKILL.md and selection-and-performance.md).
Boundary markers: None explicitly defined for raw data inputs, but the skill relies on high-level instruction-based filtering.
Capability inventory: Execution of local telemetry scripts, invocation of optional external skills (last30days), image generation, and local file writing (PNG/PDF).
Sanitization: Instructions require verifying all numerical claims against primary source documentation and performing manual fact-checking.

growth-signal-charts