The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently invokes the altitude CLI (e.g., altitude task, altitude bind, altitude notice) to manage session state, synchronize progress, and retrieve lesson data. These commands are executed locally using identifiers obtained from the environment.
[EXTERNAL_DOWNLOADS]: Instructions are provided for the user to install the @learnaltitude/cli package from the NPM registry. This package is the core tool required for the skill to function and is a resource provided by the vendor.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external sources, including CLI tool outputs (altitude task --json) and local project files (.altitude, learning/plan.md). This creates a surface for indirect prompt injection if those sources were compromised.
Ingestion points: Data is ingested from the output of shell commands and project-level metadata files (SKILL.md).
Boundary markers: The instructions explicitly direct the agent to "parse the single JSON object privately" and "never print the raw JSON, stderr, or a stack trace to the learner," which helps isolate untrusted data.
Capability inventory: The skill has the capability to execute shell commands and perform file write operations within the project directory.
Sanitization: The skill mandates that session identifiers must be passed as "safely quoted" arguments to prevent shell command injection.

begin