planning-and-task-breakdown
Warn
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: MEDIUMMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill exhibits significant metadata inconsistencies. The YAML frontmatter attributes the skill to 'addyosmani' and points to a source repository at 'github:addyosmani/agent-skills'. However, the provided author context identifies the skill's author as 'seaworld008'. This discrepancy between the internal claims and the actual publisher context represents misleading metadata regarding the skill's origin and provenance.
- [INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves reading external specifications and codebase files to generate implementation plans and task lists. This creates a surface for indirect prompt injection where instructions embedded in project files could override agent behavior.
- Ingestion points: Step 1 of the 'Planning Process' in
SKILL.mdexplicitly instructs the agent to "Read the spec and relevant codebase sections." - Boundary markers: The skill does not provide instructions to use delimiters or ignore embedded instructions when reading these external files.
- Capability inventory: The skill possesses the ability to create directories (
tasks/) and write multiple markdown files (plan.md,todo.md) to the filesystem, as well as interact with external trackers. - Sanitization: There is no evidence of sanitization or validation of the input specifications before they are processed and incorporated into the resulting task lists and plans.
Audit Metadata