The Agent Skills Directory

[SAFE]: The skill uses the Bash tool to execute 'rtk ls -t', which is a standard operation for locating project-specific session files in the current directory or a 'done/' subfolder. This command is restricted to listing files matching a specific pattern and does not incorporate unsanitized user input in a dangerous manner.- [SAFE]: The skill performs file read operations on local context and project files (e.g., manifest.yaml, tasks.md) to extract state information. This behavior is the core function of the skill and is executed within the agent's standard file access permissions.- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests and parses external markdown files (CONTEXT-.md) that could theoretically contain malicious instructions. However, this is the intended functionality of a context-loading tool. Ingestion points: Phase 1 and 2 in SKILL.md (CONTEXT-.md, manifest.yaml, and referenced artifacts). Boundary markers: None specified. Capability inventory: Bash, Read, AskUserQuestion. Sanitization: None specified. This surface is considered a low-risk inherent characteristic of context management skills.

load-context