The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the bash tool to execute a local shell script scripts/upsert-index.sh. This script updates the .context/session/INDEX.md file using standard tools like grep, sed, and printf. The script includes basic character escaping to prevent argument injection from disrupting its regex patterns. Additionally, the skill performs directory creation (mkdir -p) and file movement (mv) operations via shell commands.
[INDIRECT_PROMPT_INJECTION]: The skill synthesizes session context and task lists by analyzing conversation history, which represents a surface for indirect prompt injection.
Ingestion points: The skill explicitly reads the last 15-20 messages from the conversation history as described in Phase 2 of SKILL.md.
Boundary markers: None are present; the skill does not implement delimiters or safety warnings to ensure the agent ignores instructions that may be embedded within the conversation data being summarized.
Capability inventory: The skill possesses a broad range of capabilities, including bash, read, write, edit, and askuserquestion, as defined in its allowed-tools frontmatter.
Sanitization: No sanitization or content filtering is performed on the synthesized summaries before they are written to the filesystem as checkpoint files.

save-context