Agent skill · Audited from source
continuity-ledger
skill · parcadei/continuous-claude-v3MIT licenceAudited 13 Nov 2026 · source 79ffdab59d4c
Is it safe? No — BLOCK, grade F (36/100). We read this skill's source at commit 79ffdab59d4c and observed filesystem, network and shell access and a hard-coded secret (see the findings). The audit raised 25 findings, 12 critical or high. What it does: Agent Context Isolation
Cite
Cite this page
ModelsAtlas. “continuity-ledger (parcadei/continuous-claude-v3)”. Retrieved 13 Nov 2026. https://modelsatlas.com/skills/parcadei/continuity-ledger
Blocked. The page stays public so you can check this skill if someone recommended it to you. The audit found 12 critical or high findings, starting with “Code injection” in opc/docker/sandbox_runner.py:174. Do not install it without reading them.
Trust audit
Audit 0.4.1 · source read 13 Nov 2026
BLOCK means the audit found something that could harm you or your machine. Grade F is shared by 129 of 18,389 audited skills.
What the source does
- FilesystemDeclared1 observation
- NetworkDeclared16 observations
- ShellDeclared10 observations
- DependenciesNot pinnedSome versions float
- SecretsFoundSee the findings; values are never shown
Five audit layers
| Layer | What we check | Result | Findings |
|---|---|---|---|
L0 Provenance | Repo owner, licence, archive status, activity | WARN | 4 |
L1 Static source | Secrets, path traversal, supply chain, hidden code | FAIL | 18 |
L2 Instructions | Prompt injection, authority framing, hidden directives | FAIL | 3 |
L3 Inventory | The files and declarations match the source | PASS | 0 |
L4 Sandbox | Run in isolation and watch behaviour | SKIPPED why | 0 |
Gates applied: instruction override, no behavioural pass.
Findings (25)
HIGHCode injectioncode.eval_exec · opc/docker/sandbox_runner.py:174
exec(code, restricted_globals, local_vars)
- Why
- evaluates text as code
- Fix
- remove; use a parser or a dispatch table
HIGHCode injectioncode.eval_exec · opc/scripts/cc_math/math_base.py:526
return eval(func_str, namespace, {})- Why
- evaluates text as code
- Fix
- remove; use a parser or a dispatch table
HIGHCode injectioncode.eval_exec · opc/scripts/cc_math/math_base.py:536
return eval(lambda_str, namespace, {})- Why
- evaluates text as code
- Fix
- remove; use a parser or a dispatch table
HIGHCode injectioncode.eval_exec · opc/scripts/cc_math/pint_compute.py:284
result = eval(safe_expr, {"__builtins__": {}}, safe_namespace)- Why
- evaluates text as code
- Fix
- remove; use a parser or a dispatch table
HIGHCode injectioncode.eval_exec · opc/scripts/cc_math/scipy_compute.py:166
return np.array([eval(eq, namespace) for eq in equations])
- Why
- evaluates text as code
- Fix
- remove; use a parser or a dispatch table
HIGHNetwork egressnet.tls_off · opc/scripts/core/db/embedding_service.py:457
self._client = httpx.AsyncClient(timeout=30.0, verify=False)
- Why
- certificate verification is disabled
- Fix
- leave verification on
HIGHHard-coded secretssecret.db_uri · .claude/hooks/dist/shared/db-utils-pg.js:28
'…';
- Why
- No further detail recorded for this rule.
- Fix
- —
HIGHHard-coded secretssecret.db_uri · .claude/hooks/src/shared/db-utils-pg.ts:37
'…';
- Why
- No further detail recorded for this rule.
- Fix
- —
HIGHHard-coded secretssecret.db_uri · opc/pyproject.toml:164
runner = "…"
- Why
- No further detail recorded for this rule.
- Fix
- —
HIGHHard-coded secretssecret.db_uri · opc/scripts/claude_spawn.py:199
"…"
- Why
- No further detail recorded for this rule.
- Fix
- —
HIGHPrompt injectionprompt.credential_read · .claude/skills/hook-developer/SKILL.md:205
**Purpose:** Initialize session, load context, set environment variables.
- Why
- asks the agent to read credentials
- Fix
- —
HIGHPrompt injectionprompt.override · README.md:1120
**Windows users:** Use PowerShell (as Administrator or with Developer Mode enabled):
- Why
- asks the agent to drop prior instructions or safety
- Fix
- remove the instruction
MEDIUMInventory / provenanceinv.symlink · .claude/scripts/mcp
.claude/scripts/mcp
- Why
- link not followed
- Fix
- —
MEDIUMCode injectioncode.dynamic_import · opc/scripts/setup/math_features.py:35
__import__(pkg)
- Why
- No further detail recorded for this rule.
- Fix
- —
MEDIUMNetwork egressnet.raw_ip · .claude/hooks/dist/compiler-in-the-loop.js:14
const LMSTUDIO_BASE_URL = process.env.LMSTUDIO_BASE_URL || 'http://127.0.0.1:1234';
- Why
- No further detail recorded for this rule.
- Fix
- —
MEDIUMNetwork egressnet.raw_ip · .claude/hooks/dist/compiler-in-the-loop.mjs:6
var LMSTUDIO_BASE_URL = process.env.LMSTUDIO_BASE_URL || "http://127.0.0.1:1234";
- Why
- No further detail recorded for this rule.
- Fix
- —
MEDIUMNetwork egressnet.raw_ip · .claude/hooks/src/compiler-in-the-loop.ts:17
const LMSTUDIO_BASE_URL = process.env.LMSTUDIO_BASE_URL || 'http://127.0.0.1:1234';
- Why
- No further detail recorded for this rule.
- Fix
- —
MEDIUMNetwork egressnet.raw_ip · opc/scripts/loogle_search.py:24
SERVER_URL = "http://127.0.0.1:8340"
- Why
- No further detail recorded for this rule.
- Fix
- —
MEDIUMNetwork egressnet.raw_ip · opc/scripts/loogle_server.py:197
print(f"Loogle server listening on http://127.0.0.1:{PORT}")- Why
- No further detail recorded for this rule.
- Fix
- —
MEDIUMHard-coded secretssecret.db_uri · .claude/rules/cross-terminal-db.md:7
DATA…://…@localhost:5432…
- Why
- No further detail recorded for this rule.
- Fix
- —
MEDIUMSupply chainprompt.pipe_to_shell · .claude/skills/prove/SKILL.md:25
curl https://raw.githubusercontent.com/leanprover/elan/master/elan-init.sh -sSf | sh
- Why
- No further detail recorded for this rule.
- Fix
- —
LOWInsecure cryptocrypto.weak_hash · .claude/hooks/dist/__tests__/daemon-client.test.js:29
const hash = crypto.createHash('md5').update(projectDir).digest('hex').substring(0, 8);- Why
- No further detail recorded for this rule.
- Fix
- —
Audit 0.4.1 · source sha 79ffdab59d4c · 13 Nov 2026Report an issue or request a re-scan
Provenance
Who published it, and what the verdict is pinned to
- Publisherparcadei GitHub userOwner of the repository this skill is read from
- Source folderparcadei/continuous-claude-v3The exact folder we audited
- LicenceMITAs the repository states it
- Commit pinned for this verdict79ffdab59d4cA new audit runs when the source changes
- Repository statusNot archivedLast push 290 days ago
- Sandboxed runNot runSkills are read, not executed; layer L4 is skipped
What it tells the agent
Full source on GitHub ↗Its description: Agent Context Isolation
---
name: agent-context-isolation
description: Agent Context Isolation
user-invocable: false
---
# Agent Context Isolation
Prevent agent output from polluting the main context window.
## Rules
### 1. Use Background Agents with File-Based Coordination
```
# RIGHT - background agent writes to file, main reads file
Task(subagent_type="...", run_in_background=true, prompt="... Output to: /path/to/file.md")
# WRONG - foreground agent dumps full transcript into main context
Task(subagent_type="...", run_in_background=false)
```
Background agents with `run_in_background=true` isolate their context. Have them write results to files in `.claude/cache/agents/<agent-type>/`.
### 2. Never Use TaskOutput to Retrieve Results
```
# WRONG - dumps entire transcript (70k+ tokens) into context
TaskOutput(task_id="<id>")
TaskOutput(task_id="<id>", block=true)
# RIGHT - check expected output files
Bash("ls -la .claude/cache/agents/<agent-type>/")
Bash("bun test") # verify with tests
```
TaskOutput returns the full agent transcript. Always use file-based coordination instead.
### 3. Monitor Agent Progress via System Reminders
```
# System reminders come automatically:
# "Agent a42a16e progress: 6 new tools used, 88914 new tokens"
# To detect completion:
# - Watch for progress reminders to stop arriving
# - Poll for expected output files: find .claude/cache/agents -name "*.md" -mmin -5
# - Check task output file size growth: wc -c /tmp/claude/.../tasks/<id>.output
```
**Stuck agent detection:**
1. Progress reminders stop arriving
2. Task output file size stops growing
3. Expected output file not created after reasonable time
### 4. Verify with Tests, Not Output
After agent work:
1. Run the test suite directly: `bun test`
2. Report pass/fail counts
3. Only investigate failures if tests fail
### 5. File-Based Agent Pipeline Pattern
```
Research agent → .claude/cache/agents/oracle/output.md
↓
Plan agent → .claude/cache/agents/plan-agent/output.md (reads research)
↓
Validate agent → .claude/cache/agents/validate-agent/output.md (reads plan)
↓
Implement agent → src/module.ts (reads validated plan)
```
Each agent reads the previous agent's file output, not TaskOutput.
## Why This Matters
Agent context isolation preserves the main conversation's context budget. Reading agent outputs via TaskOutput floods context, causing:
- Mid-conversation compaction
- Lost context about user's original request
- Repeated explanations needed
## Source
- Session where TaskOutput flooded 70k+ tokens into main context
- Session 2026-01-01: Successfully used background agents with file-based coordination for SDK Phase 3Excerpt, first 2,714 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.
Install
Pinned to the audited commit 79ffdab
uv run python -m runtime.harness scripts/perplexity_ask.py \ pip install agentica uv add agentica uv run pytest tests/unit/ -v --tb=short -q uv run pytest tests/unit/ --cov=src --cov-report=term-missing uv run pytest tests/integration/ -v --tb=short
git clone https://github.com/parcadei/continuous-claude-v3 git -C continuous-claude-v3 checkout 79ffdab59d4c
Commands as the repository documents them: shown, never run by us. Our verdict covers commit 79ffdab59d4c only.
Works in which agents?
What its own text says, not a test
| Agent | Status | More |
|---|---|---|
| Claude Code | Mentioned in its text, not tested | Skills for Claude Code · About Claude Code |
| GitHub Copilot | Mentioned in its text, not tested | Skills for GitHub Copilot · About GitHub Copilot |
| Cursor | Mentioned in its text, not tested | Skills for Cursor · About Cursor |
Audit history
Every audit, every source change
- ✕BLOCK F 36 / 100
first audit · source
79ffdab59d4c
README badge
For maintainers · shows the latest audit
[](https://modelsatlas.com/skills/parcadei/continuity-ledger/)
The badge shows our verdict and grade for this skill, and links back to this page. It changes when the audit does.
Similar skills
parcadei/continuous-claude-v3
Transform session learnings into permanent capabilities (skills, rules, agents). Use when asked to "improve setup", "learn from sessions", "compound learnings", or "what patterns should become skills".
parcadei/continuous-claude-v3
Braintrust tracing for Claude Code - hook architecture, sub-agent correlation, debugging
parcadei/continuous-claude-v3
Workflow orchestrator that chains existing skills for feature development
parcadei/continuous-claude-v3
Claude Code CLI commands, flags, headless mode, and automation patterns
parcadei/continuous-claude-v3
Create git commits with user approval and no Claude attribution
parcadei/continuous-claude-v3
A complete skill for E2E testing
Add this badge to your README
[](https://modelsatlas.com/skills/parcadei/continuity-ledger/)
<a href="https://modelsatlas.com/skills/parcadei/continuity-ledger/"><img src="https://modelsatlas.com/skills/parcadei/continuity-ledger/badge.svg" alt="ModelsAtlas: BLOCK F"></a>
Questions about continuity-ledger
What does the continuity-ledger skill do?
Is continuity-ledger safe to install?
What can continuity-ledger access on my machine?
Which assistants does continuity-ledger work with?
How current is this page?
Sources
- github.com/parcadei/continuous-claude-v3 — the audited source at commit 79ffdab59d4c and every finding aboveObserved · 13 Nov 2026
Spotted something wrong? Report a correction · Methodology
Keep going in your AI
Opens with this page as the source. We add nothing else to the prompt.
- What does the continuity-ledger skill do?
- Is continuity-ledger safe to install?
- What can continuity-ledger access on my machine?