The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted conversation history and user-provided project names to generate markdown summaries and update a central memory log, creating an attack surface where malicious instructions in the conversation could influence agent behavior.
Ingestion points: The conversation topic and history are used to 'inferred project name' and 'Summarize' content in SKILL.md.
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the text being summarized, increasing the risk of the agent following instructions found within the data.
Capability inventory: The skill has the capability to read, write, and delete files and create directories within the local vault filesystem.
Sanitization: The instructions do not specify any validation or sanitization of the conversation content before it is written to the vault or appended to the memory.md file.
[COMMAND_EXECUTION]: The skill performs file management operations including directory creation and file deletion.
Evidence: The /tldr delete operation identifies files using a glob pattern (*-{project-name}-tldr.md). As noted in the 'Gotchas' section, this can lead to unintended matches and potential accidental deletion of unrelated project files if project names are subsets of others (e.g., 'api' matching 'payment-api').
Evidence: The skill automatically creates project subfolders and writes markdown files to paths determined by user input and conversation context.

tldr