The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's initialization preamble executes several local binary scripts such as econstack-update-check and econstack-learnings-read. These files reside in ~/.claude/skills/econstack/bin/, which is outside the skill's own directory, making their internal logic unverifiable from the skill's source code.
[REMOTE_CODE_EXECUTION]: The instruction eval "$(~/.claude/skills/econstack/bin/econstack-slug)" uses eval on the output of a script. This is a dangerous pattern that executes the script's output directly as shell commands, which could lead to arbitrary code execution if the script is compromised or produces malicious output.
[EXTERNAL_DOWNLOADS]: The skill recommends that the user install the debtkit R package directly from a third-party GitHub repository (charlescoverdale/debtkit) which is not an official or well-known trusted source.
[EXTERNAL_DOWNLOADS]: The skill suggests cloning an external data repository (econstack-data) from a third-party GitHub user (charlescoverdale) to enable full parameter support, posing a supply chain risk.
[DATA_EXFILTRATION]: The skill implements an update check mechanism and a 'learnings' logging system (econstack-learnings-log) that tracks user insights and project details. This implies network activity for updates and the collection of potentially sensitive contextual information from the agent's session.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests large volumes of data from external government APIs and local files to generate its briefings.
Ingestion points: Reads core fiscal indicators from various agencies including ONS, HMRC, HMT, OBR, BEA, Treasury, CBO, OMB, ABS, and PBO. It also ingests local fact files and parameters from ~/econstack-data/parameters/.
Boundary markers: While the skill uses a structured output template and enforces citation discipline, it lacks explicit boundary markers or instructions to the agent to ignore any embedded commands within the fetched data.
Capability inventory: The skill has access to Bash for executing shell commands and fetching data, and Write for creating report files.
Sanitization: There is no evidence of input validation or sanitization for the data retrieved from external sources before it is processed by the agent.

fiscal-briefing