The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/session-export-nocode.sh uses exec bash to launch a secondary export script (export-direction.sh). The installation instructions in SKILL.md also direct the user to run commands using npx and bash.
[DYNAMIC_EXECUTION]: The launcher script scripts/session-export-nocode.sh determines the path to the base runner script at runtime by iterating through several potential installation directories (e.g., ~/.agents/skills, ./.claude/skills). This dynamic path resolution is used to load and execute code.
[DATA_EXFILTRATION]: The skill's primary purpose is to export session trajectories. While the export is intended for local storage, the skill explicitly provides a mechanism (--allow-credentials) to include sensitive session data that would otherwise be filtered out. The instructions guide the agent to skip user confirmation by default using the --yolo flag unless specific confirmation phrases are encountered.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted session history data which may contain embedded instructions designed to influence the agent's behavior during the export process.
Ingestion points: Previous session trajectories and interaction logs stored on the local system.
Boundary markers: The skill does not define or use specific boundary markers to separate session content from processing instructions.
Capability inventory: The skill has the ability to execute shell scripts via bash and write data to the filesystem during the export process.
Sanitization: The skill uses a keyword exclusion list (exclude_keywords in direction.json) to filter which sessions are exported, but it does not perform sanitization of the content within those sessions to prevent prompt injection.

session-export-nocode