The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze Google Tag Manager container exports in JSON format. This represents an indirect prompt injection surface where a malicious actor could embed instructions within tag names or triggers in a container to influence the agent's audit report.
Ingestion points: Processes GTM container JSON exports provided by the user (SKILL.md, EXAMPLES.md).
Boundary markers: The skill does not specify explicit delimiters or "ignore embedded instructions" warnings for the untrusted container data.
Capability inventory: The skill is primarily instructional but mentions a related gtm-implementer agent for writing changes via API.
Sanitization: No specific sanitization or validation of the input JSON structure or content is described.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to download its source code and other marketing skills from a remote repository.
Evidence: git clone https://github.com/thatrebeccarae/claude-marketing.git (SKILL.md).
Context: The repository belongs to the skill's author, representing a legitimate vendor resource.
[METADATA_POISONING]: The skill includes futuristic dates in its metadata and reference materials which could be misleading.
Evidence: updated: 2026-03-18 in SKILL.md and references to TCF v2.3 compliance as of February 2026 in REFERENCE.md.
Context: While likely a stylistic choice or error, inaccurate metadata is technically deceptive.

google-tag-manager