The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses the local file ~/.config/catalyst-cloud/customer.json to verify the presence of authentication credentials such as API keys or OAuth refresh tokens. Reading sensitive credentials from the filesystem represents a significant data exposure surface.\n- [COMMAND_EXECUTION]: The skill spawns the catalyst-skills CLI using child_process.spawnSync. It passes user-supplied inputs, such as ticket identifiers and justification strings, as arguments to the tool. On Windows, the process is executed within a shell environment, which could be leveraged for command injection if input is not rigorously validated.\n- [EXTERNAL_DOWNLOADS]: The skill uses npx to execute the @catalyst-cloud/catalyst-skills package, which may involve downloading code from the NPM registry at runtime. This package is part of the vendor's own infrastructure and is necessary for the skill's operations.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes ticket summaries, history, and failure reasons retrieved from Catalyst Cloud. Adversarial content embedded in these fields could potentially influence the agent's behavior during the ticket release process.\n
Ingestion points: Ticket metadata and execution history are read via CLI commands in scripts/unstick.mjs.\n
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat external ticket data as untrusted or to ignore instructions contained within it.\n
Capability inventory: The skill has the capability to execute commands and perform cloud-side modifications through the release function.\n
Sanitization: The skill does not perform sanitization of the ticket data before it is presented to the agent for analysis.

unstick