The Agent Skills Directory

[PERSISTENCE]: The skill implements a hook install command that modifies the agent's configuration file at .claude/settings.json. This adds a SessionStart hook, ensuring the signatories-axi home command executes automatically whenever a new agent session begins.
Evidence: The install function in scripts/signatories-axi.mjs calls computeSessionStartHookUpdate and writeSettings to persist a command in the agent's settings.
[INDIRECT_PROMPT_INJECTION]: The skill enables the export of feedback comments from an external API for processing by an LLM, creating a vulnerability to indirect prompt injection.
Ingestion points: The feedback export command in SKILL.md and scripts/signatories-axi.mjs fetches document comments from the Signatories API.
Boundary markers: The skill does not provide or enforce delimiters or safety warnings for the LLM when processing this external content.
Capability inventory: The skill has permissions for network operations, local file writes, and modifying agent configuration files.
Sanitization: No sanitization or filtering is performed on the ingested comments before they are presented to the agent.
[DYNAMIC_EXECUTION]: The skill dynamically constructs shell command strings and writes them into the agent's session configuration to enable the automated dashboard functionality.
Evidence: The variable PROJECT_HOOK_COMMAND in scripts/signatories-axi.mjs contains a command string that is written to .claude/settings.json during the installation process.
[COMMAND_EXECUTION]: The skill executes local shell commands to determine the environment state, such as identifying the project root.
Evidence: The gitRoot function in scripts/signatories-axi.mjs uses execSync("git rev-parse --show-toplevel") to interact with the underlying system.

signatories-axi