The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the gh (GitHub CLI) command to automate the creation of bug reports in the tenstorrent/tt-lang, tenstorrent/tt-metal, and tenstorrent/tt-xla repositories.
[DATA_EXFILTRATION]: The skill reads specific compiler artifacts, namely /tmp/ttlang_initial.mlir and /tmp/ttlang_final.mlir, to include diagnostic context in reports sent to GitHub. This behavior is restricted to the specific files necessary for the skill's stated purpose.
[PROMPT_INJECTION]: The skill ingests untrusted user data to populate the bug report fields, presenting an indirect prompt injection surface.
Ingestion points: User-provided input is captured via the $ARGUMENTS variable in SKILL.md.
Boundary markers: No specific boundary markers or instructions are provided to the agent to ignore potentially malicious instructions within the user's problem description.
Capability inventory: The skill uses gh for network-based issue creation and reads files from the local filesystem.
Sanitization: The skill uses a quoted heredoc (<<'EOF') in the bash script template, which prevents local shell expansion of the issue body but does not filter the content itself.

tt-bug-report