The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses instructions to enforce autonomous behavior with reduced user oversight, such as "minimal interruption," "Do not pause for non-critical confirmation," and "Execute tasks... without stopping."
[COMMAND_EXECUTION]: The skill executes various local scripts and tools as part of the project lifecycle, including node scripts/check-phase-entry.mjs and pnpm tasks (pnpm check:truth-gates, pnpm typecheck, pnpm lint, pnpm test).
[PROMPT_INJECTION]: Trigger conditions include Uzbek phrases ("to'xtama", "bir passda tugat", "oxirigacha bajar") to activate the autonomous mode, which can bypass simple English-language instruction filters.
[INDIRECT_PROMPT_INJECTION]: The skill processes project-specific files to determine the execution loop and state, which serves as an injection surface if these files are influenced by untrusted sources.
Ingestion points: HANDOFF.md, _memory/progress.md, _planning/phase-N/README.md, and docs/TASKS.md are read to guide execution.
Boundary markers: None identified in the instructions for delimiting content from these files.
Capability inventory: The skill has the capability to execute shell commands (node, pnpm) based on the tasks identified in the ingested files.
Sanitization: No sanitization or validation of the content within the ingested files is mentioned before the agent acts upon them.

autonomous-delivery