Security audit
Security checks for vulnerabilities and agentic risk
The skill is not proven malicious, but it asks users to run a mutable remote installer that adds a persistent Claude hook capable of rewriting future Bash commands and using broad local history/context data.
Review this carefully before installing. Only install if you trust the publisher and are comfortable with a persistent Claude hook that can alter Bash commands before execution. Prefer a version-pinned, inspectable, verified installer, review the exact `~/.claude/settings.json` changes, keep backups, avoid `precc ingest --all` on sensitive histories, and confirm there is a clear uninstall and data-deletion path.
SKILL.md:101Unpinned Remote Installer Is Executed Directly Through Bash
SKILL.md:105Installer Establishes a Persistent PreToolUse Command-Interception Hook
SKILL.md:52Broad Session-History and Agent-Context Access Lacks Documented Scope Controls
The skill declares access to ~/.claude/settings.json, which is a sensitive agent configuration file controlling hooks and runtime behavior. Modifying agent config can establish persistent execution, alter trust boundaries, or silently intercept future tool invocations, making this especially sensitive in a skill that installs a hook.
config:
- ~/.local/share/precc/history.db
- ~/.local/share/precc/heuristics.db
- ~/.claude/settings.json
env:
- PRECC_LICENSE_KEY
primaryEnv: PRECC_LICENSE_KEY
Fetching and executing an external install script from a remote domain introduces a high-risk supply-chain and remote-code-execution vector. Even if the script currently verifies downloaded binaries, the initial script itself runs with the user's privileges before any independent trust validation is possible.
curl -fsSL https://peria.ai/install.sh | bash
precc init
The '| bash' chain removes the user's opportunity to inspect the fetched content and immediately executes whatever the remote endpoint returns. In a skill that also installs hooks and edits agent configuration, this chaining meaningfully increases the blast radius because compromise can lead to both arbitrary code execution and durable agent persistence.
curl -fsSL https://peria.ai/install.sh | bash
precc init
This section explicitly states that installation configures a PreToolUse hook in ~/.claude/settings.json. Because hooks execute during future agent operations, this is effectively persistence inside the agent environment and could be abused to monitor, modify, or redirect subsequent commands and data flows.
precc init
The install script downloads a platform-specific binary from GitHub Releases, verifies its SHA256 checksum, and places it in `~/.local/bin`. It then configures a PreToolUse hook in `~/.claude/settings.json`.
## Live Status Line
The statusLine command configuration in ~/.claude/settings.json causes a local binary to run repeatedly within the agent UI context. While presented as observability, any command embedded in agent config creates an execution surface that can be repurposed for persistence, data access, or misleading output if the binary is replaced or behaves unexpectedly.
To enable manually, add to ~/.claude/settings.json:
{
"statusLine": {
The skill openly states it modifies ~/.claude/settings.json to add a PreToolUse hook entry, confirming persistent changes to agent configuration. In this skill's context, that is more dangerous because the core feature is command interception and correction before execution, giving the installed component influence over future shell behavior.
## What PRECC Modifies
- **`~/.claude/settings.json`** — Adds a `PreToolUse` hook entry pointing to `precc-hook`
- **`~/.local/share/precc/`** — SQLite databases for learned failure-fix patterns and skill heuristics
- **`~/.local/bin/`** — Installs `precc`, `precc-hook`, and `precc-learner` binaries
The skill is marked user-invocable but does not define narrow activation boundaries or concrete trigger phrases. In an agent context, broad invocation semantics can cause the skill to run in unintended situations, increasing the chance that risky behaviors like installation steps, hook setup, or config modification are suggested or applied without clear user intent.
The install instructions recommend piping a remotely fetched script directly into bash without an adjacent warning or inspection step. This creates a direct code-execution path from a remote server, so compromise of the host, DNS, TLS trust chain, or release workflow could lead to arbitrary command execution on the user's machine.
No suspicious patterns detected.