The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses a helper script to download a markdown specification file from https://agentskills.io/specification.md. It also utilizes a tool to fetch documentation from https://code.claude.com/docs/en/skills#frontmatter-reference.
[INDIRECT_PROMPT_INJECTION]: The skill creates a vector for indirect prompt injection by downloading external content and directing the agent to read and follow it.
Ingestion points: The scripts/fetch-agentskills-spec.sh script downloads specification.md, which is then read into the agent's context during the skill initialization workflow.
Boundary markers: The instructions lack boundary markers or warnings to the agent to disregard potential instructions embedded within the downloaded specification.
Capability inventory: The skill has the capability to execute shell commands (mkdir, ln, chmod, curl) and use the ai-skillet command-line utility.
Sanitization: The skill performs only basic structural validation (checking for specific headers) and does not sanitize the downloaded text for malicious prompt injection patterns.
[COMMAND_EXECUTION]: The skill executes several shell commands to manage project files, including creating directories (mkdir -p), establishing relative symbolic links (ln -s), and setting executable file permissions (chmod 755) for generated scripts. These operations are within the expected scope of a development and scaffolding tool.

skill-writing