The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The software architect playbook (references/PLAYBOOK-ARCH.md) instructs the agent to recommend that the user install an external third-party skill (tt-a1i/archify) if it is not currently available. This introduces a dependency on an unverified third-party repository into the development workflow.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves ingesting and acting upon external project documentation (PRDs, user stories, and design files), which constitutes a potential vector for indirect prompt injection attacks.
Ingestion points: The skill reads PRD.md, Design.md, and various user story files as described in the playbooks within the references/ directory.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to distinguish between its core instructions and potentially malicious content within the external documents it processes.
Capability inventory: The agent has the authority to write project files (e.g., TECH-DESIGN.md, PRD.md, adrs/*.md) and coordinate with other team agents using the invoke-agent tool based on the data it ingests.
Sanitization: The instructions do not mandate any form of content validation or sanitization for the inputs retrieved from external project files before they are utilized in decision-making processes.
[NO_CODE]: The skill consists entirely of Markdown instruction files and reference playbooks, with no accompanying executable scripts or binaries.

init-agent