The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple local binaries and scripts located in ~/.claude/skills/econstack/bin/, including econstack-update-check, econstack-learnings-read, and econstack-greenbook.
[REMOTE_CODE_EXECUTION]: The skill uses eval "$(~/.claude/skills/econstack/bin/econstack-slug)". This executes the output of a local command directly in the shell environment, which is a pattern that can lead to arbitrary code execution.
[EXTERNAL_DOWNLOADS]: The skill references external resources for cloning data and installing packages, specifically https://github.com/charlescoverdale/econstack-data.git and https://github.com/charlescoverdale/greenbook. These repositories belong to an individual user and are not from established trusted organizations or well-known services.
[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by parsing untrusted project data from external markdown files via the --from argument.
Ingestion points: Project parameters, options, costs, and benefits are imported from user-provided markdown files.
Boundary markers: No delimiters or instructions are used to separate untrusted data from the skill's logic.
Capability inventory: The skill has access to Bash for command execution, Write for file creation, and the ability to invoke other skills like xlsx and docx.
Sanitization: There is no documented validation or sanitization of the imported markdown content before it is processed.

cost-benefit