The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses multiple shell commands to check for, install, and run the gjalla CLI, as well as to search for repository paths and session transcripts.- [EXTERNAL_DOWNLOADS]: Installs the 'gjalla' package from public registries using pipx or uv.- [DATA_EXFILTRATION]: The skill extracts and uploads project metadata and content to the vendor's infrastructure at gjalla.io.
Session Metadata: Step 5 extracts commit SHAs, branch names, and commit subjects from git logs and session transcripts located in ~/.claude/projects/ and ~/.codex/sessions/.
Project Memory: Step 7 gathers 'facts' from project documentation (CLAUDE.md, README.md, etc.) and git history, uploading them to the shared memory service.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from git logs and project documentation, which could contain malicious instructions.
Ingestion points: Reads from git logs via git log and git show, session transcripts in the user's home directory, and various markdown files within the project.
Boundary markers: Absent. The skill does not use delimiters to encapsulate the data being parsed or instructions to the agent to ignore embedded commands.
Capability inventory: Includes shell command execution and network operations via the gjalla CLI.
Sanitization: While the skill filters for secrets and specific headers, it does not sanitize the content to prevent command injection or prompt injection during the data gathering steps.

gjalla-onboard