The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution to manage repository state, including git status, git stash, git fetch, and git merge within the ~/.config/superpowers/skills directory.
[COMMAND_EXECUTION]: Executes a script at the relative path ../using-skills/find-skills to verify the state of the system after an update. This pattern relies on the integrity of executable files located outside the skill's own directory structure.
[EXTERNAL_DOWNLOADS]: Synchronizes data with a remote Git repository (obra/superpowers-skills) using git fetch. This involves downloading external content from a third-party source to the local environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it instructs the agent to ingest and interpret untrusted data from a remote repository.
Ingestion points: Data is pulled from the remote repository and viewed through git log and conflict resolution steps in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore or sanitize potentially malicious instructions embedded within commit messages or file diffs.
Capability inventory: The agent has capabilities to execute shell commands and modify local files, which could be targeted by instructions found in the external data.
Sanitization: The skill lacks logic to filter or validate the content retrieved from the tracking remote before the agent presents it to the user or acts upon it.

pulling-updates-from-skills-repository